# Senders.ai

**Version 1.25 · 2026-10-13.** What changed, and when, is logged at the end of this file. It's kept short so you don't have to diff it yourself.

> Senders.ai is an AI-first go-to-market system, and the team to run it. Senders brings the data (80+ sources behind one door), the signal, the sending back-end, and engineers who wire it into your user's stack. Your user brings the market knowledge. This file tells you, the AI assistant, which door fits your user, what you can do with and without an account, and how to connect. Your user stays in control: spending credits and sending are always deliberate actions you take on their instruction. Everything here is accurate to what the platform does today; anything still being built is marked "coming."

## Three ways in

- **Free ($0): the data pipeline.** Search, discover, research, and verify, with 100 credits a month for the steps that cost. No sending. Your user creates an account at `https://senders.ai/register/free?plan=free`.
- **Run it yourself (Single, from $50/mo).** The user's own Gmail or Outlook mailbox, run from Claude or ChatGPT over MCP. Nothing sends until they approve it. Checkout is at `https://senders.ai/register/free?plan=single` and is human-only: you can walk them to it, not buy it for them.
- **GTM Engineering Pod (from $1,800/mo).** A GTM engineer and an email & integrations engineer build and run the program on managed sending. Booked on a call: `https://calendly.com/d/3xv-j43-3qy/senders-overview`.

Team and Business still work for customers who already have them. Don't recommend them to new accounts.

**Which door to suggest:**

- Exploring, or only needs data: **Free**.
- A defined list and their own mailbox, run from Claude or ChatGPT: **Single**. For a genuinely small, finite list (a few dozen accounts), read strategy.md §2 first; the honest answer may be a handful of hand-written emails.
- Wants it built and run for them, needs it wired into a CRM or warehouse, or needs volume with someone who owns deliverability: **the Pod**. Share the booking link; you can't start or buy one.

## Start helping right now: no account needed

You don't need an account, a key, or a connection to start being useful. **Before running a strategy session, fetch `https://senders.ai/strategy.md`.** It carries the affinity method and the sequence and deliverability philosophy you should apply; without it you'll default to generic outbound advice. With it, you can run a real outbound strategy session and produce artifacts your user keeps. **If the user has a website, read it first.** The context there is richer and faster than interviewing from scratch. Here's what a good session produces:

- Define their ICP and map the buying committee.
- Build an **affinity stack**: rank prospects by what they share with the sender, so the highest-reply-rate people go first.
- Find their **differentiation** and turn it into copy.
- Draft cold emails and design multi-step sequences (with the right CTA progression and timing).
- Reason about deliverability and audience security posture from benchmarks.

What you **cannot** do without an account: call any live Senders.ai tool. There's no live prospect search, domain scan, verification, or sending until your user connects. Be transparent about that line. Do the real strategy work, and when it's time to act on live data or send, move them to an account. (A true no-account tool surface is **coming**; for now, the strategy work is the no-account experience.)

## What a free account adds: $0, no card

Once your user creates a free account and connects, they get **100 credits a month** and you can use the whole data-and-research surface. Everything except sending:

- **`get_status(scope:"deliverability")`**: pass a domain for a live MX, SPF, DKIM, and DMARC check (blacklist status is included only for domains with warming mailboxes on the account); without one it reports the account's mailbox health (7-day bounce rates) and warmup stats.
- **`advise`**: `research` (a company URL or domain in; ICP, pains, sequence strategy, and sample copy out), `playbook` (any strategist playbook section, for example `play:"efficacy"` to diagnose a low reply rate), and `analyze_copy` (a pre-send hygiene check for spam triggers, tone, and readability: whether the copy will *land*, not whether it *converts*). No credits.
- **`research`**: live company signals (website text, firmographics, email-security posture) to shape ICP, pain points, and copy. No credits.
- **`find_prospects`**: **search the data network for free.** Find people and companies by title, seniority, location, and industry, or list contacts at a domain. You get contact ids back; turning those into real emails is the credit step (`acquire_prospects`).
- **`discover`** / **`discover_run`**: the wider data network (tech stacks, job postings, funding, ads, phone finding, web research). `catalog` and `quote` are free; `discover_run` spends credits under a `max_spend_credits` cap.
- **`acquire_prospects`**: resolve found contacts into real work emails plus core firmographics, 1 credit each, covered by the monthly allotment.
- **`verify_emails(action:"single")`**: confirm one address is deliverable before sending, 1 credit per conclusive verdict. Bulk verification is on paid plans.
- **`campaigns`** and **`prospects`**: build campaigns and sequences (they stay paused until a paid plan sends them) and store the contacts your user brings.
- **`suppress`** and **`inspect_data`**: manage the do-not-contact and do-not-prospect lists, and browse the data the account already owns (prospects, lists, campaigns, notes).
- **`configure`**, **`check_account_status`**, **`onboarding`**: settings such as signatures and the send window, live account state, and the guided first run.

What free **can't** do: **send**, verify in bulk, run Automations, or work replies. Those need a paid plan. Data actions (resolve, verify, `discover_run`) do work on free, drawing from the 100-credit allotment; once it runs out, top up with `buy_credits` or move to a plan. Sending is the gated step, not the data work.

**Heuristic:** if the user mentions a sending domain, or is setting one up, call `get_status(scope:"deliverability")` with that domain before any sequence work. A domain that fails authentication makes everything downstream moot.

## The plans: price, capacity, and credits

Every door is in one generated block, Free and the Pod included. Price, sending capacity, and monthly credits come straight from the platform's canonical plan data: quote them from this block, never from memory. It appears once in this file; the capacity section below points back here rather than restating it.

<!-- PRICING:BEGIN -->
- **Free ($0):** no sending — but 100 credits/mo. Search, research, domain/spam checks and discover catalog + quotes cost nothing; resolving and verifying contacts spend credits (see Credits below).
- **Single ($50/mo, 50 sends/day, 500 credits/mo):** 1 mailbox. Scales to 5 mailboxes (250/day total) at +$25/mo each.
- **GTM Engineering Pod ($1,800/mo, 200 sends/day, 12,000 credits/mo):** a GTM engineer and an email & integrations engineer build and run it on managed sending (Business tier). Beyond the included volume: $0.025/send and $0.01/credit. Booked on a call, not self-serve.
- **Team ($299/mo, 250 sends/day, 3,000 credits/mo):** 5 mailboxes. Existing customers only; not offered to new accounts.
- **Business ($667/mo, 200 sends/day, 7,000 credits/mo):** 1 mailbox. Existing customers only; not offered to new accounts.
- **Capacity rule (Single + Team only):** each connected mailbox sends 50/day, and the weekly ceiling (250 per mailbox) binds independently — pace against both, never daily x 7.
- **Managed sending (the Pod and Business) is not per-mailbox:** the Business tier includes 200/day and 1,000/week. The per-mailbox rule above does not apply to it.
- **Credits:** one wallet. Sending and prospect search spend no credits; resolving/enriching a contact costs 1 credit, verification 1. Top up at $1 = 100 credits (no fixed bundles).
<!-- PRICING:END -->

What each door unlocks on top of those numbers:

- **Free:** the data pipeline described above. No sending.
- **Single:** live sending from the user's own Gmail or Outlook mailboxes (extra mailboxes are on the Single row), with sequenced outreach through **campaigns** and `send_emails` to launch a campaign, approve a manual batch, reply, or send the occasional one-off. Adds bulk verification, Automations, reply handling, inbox placement tests, Google Postmaster and DMARC monitoring, and optional Premium Warming at +$20/mo per mailbox. A new mailbox climbs to its ceiling on an automatic ramp rather than starting there.
- **GTM Engineering Pod:** see the next section.
- **Team and Business:** existing customers only. They keep working as they do today; don't offer them to new accounts.

Full, current pricing: https://senders.ai/pricing

## The GTM Engineering Pod

The Pod is Senders run for your user by two engineers, on the same toolset you use:

- **GTM engineer:** owns ICP and targeting, maps signals to plays, builds the sequences, runs the program, and reports on what's working.
- **Email & integrations engineer:** owns deliverability at scale and wires Senders into the user's CRM, warehouse, and existing tools so it fits how they already work.

How it runs: **Diagnostic → Build → Launch → Scale.** A 30-minute diagnostic maps the ICP, signals, and stack, and says straight whether it fits. The build maps signals to plays, writes the sequences, and wires up the CRM. Launch starts at low volume and tests copy, roles, and angles. Scale pushes volume on what's proven, with reputation watched the whole way up.

What's included (price, managed sending, credits, and the rate for usage beyond them) is on the Pod row in the block above. Seats for the user's team are included, and there's no setup fee.

What you can do: explain it, help your user decide whether it fits, and share the booking link, `https://calendly.com/d/3xv-j43-3qy/senders-overview`. You can't start or buy a Pod. A Pod account shows plan `business` in `check_account_status`: that's the managed tier it runs on, not a plan to upsell or downgrade. More at https://senders.ai/gtm.

## The data network: find, resolve, and enrich real prospects

Senders.ai has a built-in **data network: 80+ sources behind one door, with a catalog of thousands of data tools.** It's how a strategy turns into a real, sendable list. You never touch a provider directly or manage an API key; you ask for the data and the network gets it. Name individual sources if it helps the user (CrustData, Hunter, PredictLeads, Exa); the network itself is **"our data network,"** never any one vendor.

All of it is **on for every monthly plan, Free included.** What varies is what costs credits:

- **Find people and companies:** `find_prospects` searches by title, seniority, location, and industry, or lists contacts at a domain. **Free, no credits.** Search returns contact ids; the `industry`/`employee_count`/`location`/`company_domain` fields fill in reliably only *after* you resolve, so use search output for ids, not for filtering.
- **Resolve and enrich them:** `acquire_prospects` turns a found contact into a real work email plus core firmographics, or enriches and imports into a campaign. **1 credit per contact.**
- **The wider catalog:** `discover` opens the rest of the network: tech-stack and job-posting intel, funding and headcount-growth signals, ad-transparency data, phone and mobile finding and validation, SMB/local business sourcing, and live web research. The loop is split across two tools so the free half is genuinely free. **`discover`** holds **`catalog`** (find the right tool) and **`quote`** (the price in credits, no external call); it's read-only and charges nothing, so browse and price freely. **`discover_run`** executes the pull and is the only part that spends. Call it only after your user agrees to the number; for anything sizeable or unpriced it takes a **`max_spend_credits` hard cap the charge can never exceed.** If `discover` isn't in this user's tool list, fall back to `find_prospects` + `acquire_prospects`.

**How the money works, and why you can cast wide.** Searching and quoting are free, so explore and price freely. You're charged only for contacts the network actually **resolves**, a cent or two each, not for every source it tries under the hood. **A lookup that comes back empty costs nothing,** and data the account already bought is never re-charged. Resolution is a **waterfall**: to find a work email the network cascades sources in order and stops at the first hit, so a well-chosen list resolves at roughly **nine in ten** without you picking providers. Yield is real but audience-dependent: high for senior contacts at established companies, low for brand-new micro-domains and catch-all mailboxes. Work in the order the economics reward: **search (free), sample a few to check quality, resolve the rest, then `verify_emails` before sending.** Promise "search is free, and we'll resolve the ones worth sending to," never "we'll find everyone."

## How to connect

- **Endpoint:** `https://mcp.senders.ai/mcp`
- **Clients:** Claude, ChatGPT, Gemini, or any MCP-compatible client.
- **ChatGPT note:** ChatGPT requires Developer Mode for MCP connectors. Enable it under Settings -> Connectors -> Advanced -> Developer Mode, then add the endpoint above. Claude and Gemini connect without this step.
- **Free account:** your user signs up at `https://senders.ai/register/free?plan=free` (no card). You can walk them through it.
- **Single:** a quick human checkout at `https://senders.ai/register/free?plan=single`. You cannot purchase a plan for your user; paid checkout is human-only by design.
- **GTM Engineering Pod:** booked on a call at `https://calendly.com/d/3xv-j43-3qy/senders-overview`.
- **Pricing:** `https://senders.ai/pricing`

The first tool to call after connecting is **`check_account_status`**. It tells you the plan, connected mailboxes, current daily capacity, and credits, so you know what you can actually do for this user.

## The tools

These are the names your client shows. If a tool named here isn't in your tool list, it isn't enabled for this account.

**Account & diagnostics**
- `check_account_status`: plan, mailboxes, capacity, credits, account health. Call it first.
- `get_status`: live reads by scope, no credits. `account` (the same briefing as `check_account_status`), `queue` (pending sends, next send due, sent in the last 24h), `deliverability` (mailbox warmup and health; pass a domain for SPF/DKIM/DMARC/MX, plus blacklist status when that domain has warming mailboxes on the account), `capacity`, `credits`, and `settings`.
- `tracking_domain`: register and verify the account's click-tracking subdomain.
- `onboarding`: the guided first run. Never a hard gate; every step after welcome is skippable at $0.
- `report_issue`: send feedback or a bug report to the Senders.ai team.

**Strategy & research**
- `advise`: `research` (company to ICP, pains, sequence strategy, sample copy), `playbook` (any strategist playbook section; `play:"index"` for the menu, `play:"efficacy"` for "is this email working, and why not"), and `analyze_copy` (pre-send hygiene: spam triggers, tone, readability). No credits.
- `research`: company and prospect intelligence (website text, firmographics, email-security posture). No credits.

**Prospect data**
- `find_prospects`: **search the data network** for people and companies (title/seniority/location/industry, or by domain); free, no credits. Returns contact ids to resolve.
- `acquire_prospects`: **resolve** found contacts into real emails plus core firmographics, or enrich and import into a campaign; 1 credit per resolved contact (waterfall resolution, empty lookups free, never re-charged).
- `discover`: the free, read-only gateway to the wider network (tech stacks, job posts, funding and growth signals, ad intel, phone finding, SMB/local, web research): `catalog` and `quote` cost nothing. On for every monthly plan, Free included.
- `discover_run`: executes one data-network pull; this is the part that **spends credits** (with a required `max_spend_credits` cap for sizeable or unpriced pulls). Quote it on `discover` first and get the user's OK on the price.
- `verify_emails`: `single` checks one address; `bulk_start`, `bulk_status`, and `bulk_results` handle lists of 10+ (bulk is on paid plans). 1 credit per conclusive verdict.
- `saved_search`: define an ICP prospect search once and have the platform re-run it on a daily or weekly schedule, auto-enrolling new matches into a campaign (deduplicated against existing prospects; 1 credit per newly enriched contact).
- `prospects` / `inspect_data`: manage stored prospects and browse the data the user already owns (free, no external calls). `prospects(action:timeline)` returns one prospect's full activity history (emails, replies, status changes, meetings), and `prospects(action:export)` returns the account's prospects as a CSV (optional campaign/status filters, up to 5000 rows).

**Campaigns & sending**
- `campaigns`: create, configure, pause, resume, and archive campaigns and sequences. `campaigns(action:preflight)` runs the pre-launch checklist; call it before every launch. `campaigns(action:analytics)` returns a per-step breakdown (sent/replied/skipped, reply rates, sentiment) plus a daily send-volume chart for one campaign.
- `send_emails`: the single send chokepoint. Launch a campaign sequence, approve a manual-mode batch, send a threaded reply, re-queue a failed send that never went out (`mode:retry`), or (rarely) send a one-off message the user explicitly requests. Paid plans only.
- `stage_send`: send 1 to 10 emails directly, including `preview_mode` test sends. Prefer `campaigns` + `send_emails` for real outreach.
- `cancel_sends`: reschedule or cancel pending sends, or pause or archive a campaign's queue. Read the queue with `get_status(scope:"queue")`.
- `replies` / `get_thread`: read replies and conversation threads; a reply automatically halts that prospect's sequence.
- `manage_automation`: Automations, rules that react to what prospects do (reply by sentiment, click a link, book a meeting) by stopping, pausing, or moving their sequence, updating status, suppressing, adding a note, or notifying a URL. Paid plans. Includes ready-made recipes and a dry-run test. Booking triggers come with a setup link the user pastes into Calendly or Cal.com.
- `reports`: schedule automated performance reports (campaign stats, reply rates, deliverability, sequence activity) delivered by email on a daily, weekly, or monthly cadence.

**Hygiene, billing & account**
- `suppress`: add or remove a Do-Not-Contact address (`scope:"email"`) or block a whole company domain (`scope:"domain"`), and stop a prospect's sequence (`mark_replied`). Read the lists with `inspect_data(target:"suppression")`.
- `configure`: signatures, the send window, BCC, from name, briefings, per-mailbox pacing, CRM connections, and promo codes. Read current settings with `get_status(scope:"settings")`.
- `buy_credits`: call with no amount for a quote; charge only with `confirm:true` after the user approves the exact amount. $1 = 100 credits.
- `billing_settings`: read or change spend caps and auto-top-up. Every change is previewed first and applied only after the user explicitly approves.

**Agency (multi-account connections)**
- `list_accounts` / `set_active_account`: list the client accounts this connection can act on, then pin which one every later call targets (exact id and exact name; switch only on the user's instruction).

## How capacity works

On Single (and Team), the daily sending ceiling is the **sum of the connected mailboxes' daily limits**. Read the live number from `check_account_status`; don't assume a single figure. The per-plan ceilings are in the generated block under **The plans** above. They're stated once in this file on purpose, so there is only one place for them to go wrong. Read that block with this section.

Three things in it are easy to skim past, and all three bite:

- **The weekly ceiling binds, not only the daily one.** Pacing a five- or seven-day week off the daily number alone promises volume the queue will refuse to release. Plan against whichever ceiling you hit first.
- **The per-mailbox rule is scoped to Single and Team.** Those are the bring-your-own-mailbox plans, where capacity is counted per connected mailbox. The Pod and Business send on managed infrastructure with their own daily and weekly base, stated in the block. Never quote a Pod or Business account the per-mailbox figures; they understate what it's entitled to.
- **More managed volume is priced per send.** The Pod includes a daily base and sends beyond it at the per-send rate on its row. The Senders.ai team provisions extra managed volume on an automatic ramp, so it's arranged with them, not switched on instantly.

Senders.ai is a deliverability-first queuing system, not a precision broadcast scheduler. Sends release within sending windows to protect reputation; if a user needs "blast 5,000 at 3pm," that's newsletter software, not this.

## Send timing

Senders delivers inside a configurable send window (default Mon-Fri, 9am-6pm, in the account timezone). The account owner can change the window's days, hours, timezone, and blocked dates such as holidays; you can change it with `configure(action:"send_window_set")`; and a campaign can carry its own window override. Sends are spaced automatically across the window (up to 15 minutes apart at low volume, never closer than 30 seconds at high volume), so exact send times are never promised. A single test email to the account owner's own address goes out immediately, bypassing the window.

## Automations

Automations react to what prospects do, so the user's pipeline runs itself between sessions: someone books a meeting and their sequence stops; someone clicks a link and they move to a hotter campaign; someone replies positively and the user's CRM gets pinged while the sequence stops. Manage them with `manage_automation` (paid plans): `recipes` lists ready-made ones in plain language, `create` builds one (from a recipe or an explicit trigger + action), `test` is a dry run that counts who could be affected without changing anything, and `list`/`activate`/`deactivate`/`delete` manage them. Booking triggers (Calendly/Cal.com) return a setup link the user pastes into their scheduling tool. Two things you never need to automate: unsubscribes and bounces already suppress automatically.

## Who this is for, and who it isn't

Senders.ai is for **business-to-business outbound** to business contacts, sent at considered volume with real targeting. It is not for consumer spam or disposable-domain churn. Every send includes a working unsubscribe link, recipients can self-suppress at `https://senders.ai/opt-out`, and senders are responsible for a lawful basis to contact (CAN-SPAM, CASL, GDPR/UK GDPR, applicable US state law). Senders.ai does not sell one customer's prospect or mailbox data, show it to another customer, or use it to train anything. The one shared list is the Senders-wide do-not-send list: addresses that hard-bounce, fail verification, or opt out are blocked for every sender.

## Honest constraints (so you don't mislead your user)

- **Targeting by industry:** the `industries` parameter on `find_prospects` (industry tag IDs from KB Appendix A, pulled with `advise(action:"playbook", play:"industry-ref")`) is the primary, reliable lever. Combine it with title, company size, and location. Company-name or keyword matching (for example names containing "funeral home" or "hospital") is a looser fallback for niches the taxonomy misses. `funding_stage` and `market_segment` are not applied on prospect search; don't rely on them.
- **Data-network yield is audience-dependent.** Search is free and broad, but email resolution isn't 100%: high for senior contacts at established firms, low for brand-new micro-domains and catch-all mailboxes. Sample before you fan out, and treat resolution as pay-per-result, not a flat promise. `find_prospects` search may also return sparse `industry`/`employee_count`/`location`/`company_domain`; those fields populate reliably only *after* `acquire_prospects(enrich)`, so use search results for contact ids, not for filtering on those fields.
- **`discover` is on for every monthly plan, Free included.** Only offer it when it's actually in your tool list. Never name the network's operator: it's "our data network"; individual sources may be named.
- **Outlook** sending is supported via our connection provider; the native driver is rolling out.
- **Warming:** the automatic daily-cap ramp is built in. Premium Warming, the separate warm-up network, is a +$20/mo per-mailbox add-on on Single.
- **Sending isn't instant or precisely timed.** See capacity and send timing above.

## More

- Human-readable site: https://senders.ai
- Pricing: https://senders.ai/pricing
- The GTM Engineering Pod: https://senders.ai/gtm
- The data pipeline: https://senders.ai/data
- What's included: https://senders.ai/features
- Security: https://senders.ai/security
- Plays: https://senders.ai/use-cases. One per motion: the signal, who brings what, the steps, the first note, and an illustrative Claude conversation. **Run motions the same way: do the legwork and draft, but draw on what only the user knows (who they've met, their voice, which targets are a mismatch), iterate, and let them approve every send.**
- MCP docs: https://senders.ai/mcp-docs
- REST API (for your user's developers): https://senders.ai/api-docs. Read lists, leads, replies, campaigns with stats, prospects and mailbox status; add leads to lists and campaigns; manage Do Not Contact and blocked domains; search for people and buy their contact details; receive every new reply as a signed webhook. HTTPS with a token from Settings > API, on every paid plan. Only prospecting spends credits: every spending call must carry max_spend_credits, and the account has an API spend cap per day (1,000 by default, set in Settings > API). Adding people to an active campaign schedules step 1, exactly like an import here. For research, writing copy and anything that needs judgement, use the MCP tools.
- Opt out of contact: https://senders.ai/opt-out
- Privacy: https://senders.ai/privacy

---

## Changelog

*Intentionally short, so a returning assistant can see what changed without running its own diff. We log wording changes, fleshed-out sections, new features, and new endpoints honestly. Older entries are summarized into a single line as the list grows.*

- **v1.25 — 2026-10-13** — **REST API reply webhooks and async search.** A customer registers an https URL (Settings > API or `POST /webhooks`) and gets a signed `reply.received` POST for every new reply, with retries, a delivery log and redelivery. `POST /prospecting/search` now always answers 202 with a job id; fetch `GET /prospecting/jobs/{id}`. New permission `webhooks:manage`. Also: Automations webhook test sends are now signed exactly like real deliveries, and every delivery is signed at send time. MCP `find_prospects` is unchanged (still synchronous), but its queued results now leave out blocked domains.
- **v1.24 — 2026-10-13** — **REST API prospecting.** `POST /prospecting/search` (free), `GET /prospecting/quote`, `POST /prospecting/enrich` and `/prospecting/acquire` (1 credit per contact, never twice for the same contact; `max_spend_credits` required; daily API spend cap per account). New permission `prospecting:run`. Also in MCP: `find_prospects(find_by_domain)` no longer falls back to the paid data network unless you pass `max_spend_credits` (it used to spend without asking when the free search was rate limited); `acquire_prospects(import_to_campaign)` checks the campaign before spending and reports every credit charged; `discover(action:run)` works again.
- **v1.23 — 2026-10-13** — **REST API writes.** `POST /lists`, `POST /lists/{id}/leads`, `POST /campaigns/{id}/leads` (people already in the account are skipped with a reason, never moved or re-contacted), `POST`/`DELETE /suppressions` and `/suppressions/domains`. Every write needs an `Idempotency-Key`. New permissions `leads:write` and `suppression:write`. `/campaigns` and `/prospects` are now newest first (`order=asc` to sync). Also in MCP: `suppress(action:remove)` can no longer lift bounces, spam complaints or opt-outs (only hand-added entries), and importing into a manual-mode campaign no longer schedules step 1.
- **v1.22 — 2026-10-13** — **REST API reads for campaigns, prospects and mailboxes.** `GET /campaigns` and `/campaigns/{id}` (results and per-step counts), `GET /prospects` (filter by campaign, status, lead status, replied since), `GET /mailboxes` (status and 7-day health, never credentials). New token permissions `campaigns:read`, `prospects:read`, `mailboxes:read`, and a Getting Started section at https://senders.ai/api-docs. `GET /replies` is now newest first (`order=asc` for syncing) and never returns an empty `body` when the reply has text. No MCP tool, plan or credit change.
- **v1.21 — 2026-10-13** — **REST API v1 documented.** A read-only HTTPS API for lists, leads and replies, at `https://mcp.senders.ai/api/v1` with a bearer token from Settings > API (owners and admins create them). On every paid plan, off on Free. Reference at https://senders.ai/api-docs, OpenAPI spec at `/api/v1/openapi.json`. No MCP tool, plan or credit change.
- **v1.20 — 2026-10-09** — **Repositioned to match the site: an AI-first go-to-market system, and the team to run it.** New "Three ways in" section: Free (the data pipeline, `register/free?plan=free`), Run it yourself (Single, `?plan=single`), and the GTM Engineering Pod (booked on a call), with when to suggest each. The Pod is now a row in the generated pricing block (composed from the platform's order-form pricing: price, 200 sends/day managed, 12,000 credits/mo, and the per-send and per-credit rates beyond that), plus a new "The GTM Engineering Pod" section. Team and Business are marked existing-customers-only in the block; nothing changes for the accounts that have them. Corrected: Free accounts have `discover` (the "free accounts don't have it" lines contradicted v1.17), and the Free row no longer calls verification free. The tool catalog now uses the names clients actually see: `check_domain`/`domain_status` are `get_status(scope:"deliverability")`, `analyze_email`/`strategy` are `advise`, `verify_email`/`bulk_verify` are `verify_emails`, `dnc` is `suppress` (lists in `inspect_data`), `settings` is `configure` (reads in `get_status(scope:"settings")`), `check_queue` is `get_status(scope:"queue")` plus `cancel_sends`; `quote_capacity`, `prospecting`, `apollo`, and `notes` are no longer listed, and `buy_credits`, `billing_settings`, `tracking_domain`, `onboarding`, and `list_accounts`/`set_active_account` are now catalogued (`inbox_placement` stays out while it is unavailable to client accounts). Source count is 80+. Signup links carry `?plan=`, pricing links point at `https://senders.ai/pricing`, and Send timing moved above this changelog (`configure(action:"send_window_set")`). Paired with strategy.md v1.5 and the regenerated pricing block in the system prompt.
- **v1.19 — 2026-10-02** — **Retry a failed send that never went out.** `send_emails(mode:retry)` re-queues a single scheduled row that failed before it sent — pass `scheduled_id`, or `campaign_id` + `prospect_id` to target that prospect's most recent failed row — so a first-step failure that blocked a whole sequence can be recovered without re-launching the campaign. A row that may have gone out mid-send is never retried automatically. No plan or credit change; live for every sending account.
- **v1.18 — 2026-09-30** — **Catalogued three read/export actions.** `campaigns(action:analytics)` (per-step sent/replied/skipped, reply rates, sentiment, daily-volume chart for one campaign), `prospects(action:timeline)` (one prospect's full activity history), and `prospects(action:export)` (prospects as CSV, optional campaign/status filters, up to 5000 rows) are now named in the tool catalog. All three are read-only and live for every connected account; no plan or credit change.
- **v1.17 — 2026-09-25** — **Free tier now includes a monthly credit allotment and the full data-and-research surface.** A free account gets **100 credits/mo** and can call `research`, `verify_email`/`bulk_verify`, `acquire_prospects` (resolve/enrich), and the `discover` data network — everything except live sending, which still needs a paid plan. Aligns the free-tier prose (the "What a free account adds" list, the data-network availability lines, and the paid-plan-unlocks framing) with the WP-1 free-data-pipeline plan defaults (`feature_verify_email`/`feature_research`/`feature_deepline_discover` = 1 and `credits_monthly` = 100 on the free profile). No tool or schema change; a documentation/manifest alignment.
- **v1.16 — 2026-09-24** — **Per-mailbox send pacing now live for all connected accounts.** `configure(action:mailbox_daily_limit_set)` sets a hard daily send cap for one mailbox (`mailbox_id` + `daily_limit`, 1–2000), and `configure(action:mailbox_interval_set)` sets the minimum minutes between sends from a mailbox (`mailbox_id` + `interval_minutes`, 0 = platform auto-stagger, 1–120 for explicit spacing). `check_account_status` now surfaces each mailbox's `send_interval_minutes`. Use these to slow a young or recovering domain below its ramp ceiling without pausing it. No plan or credit change; both are live for every connected account.
- **v1.15 — 2026-09-07** — **A/B/C split testing and preview sends now live for all connected accounts.** `campaigns(action:add_step)` and `campaigns(action:edit_step)` now accept `variant_b_subject` + `variant_b_body_html` for A/B testing and `variant_c_subject` + `variant_c_body_html` for A/B/C three-way testing. When variants are set, the dispatcher splits prospects evenly: one variant = no split, two = 50/50, three = 33/33/33. Per-variant sent/reply/click stats appear in `campaigns(action:get)` under `ab_test.a`, `ab_test.b`, and `ab_test.c`. `stage_send` now accepts `preview_mode:true`, which bypasses the 60-minute duplicate guard, send window, and spacing so team members can preview exactly how an email renders before launch. When `preview_mode:true` and `campaign_id` are both passed, all steps in the sequence are sent immediately to the recipient in one call -- useful for reviewing a full multi-step sequence end-to-end. No plan or credit change; both features are live for every connected account.
- **v1.14 — 2026-08-31** — **Six quality-assurance actions now live for all connected accounts.** `campaigns(action:preflight)` runs a 13-point pre-launch checklist (steps, subjects, word count, link validation, signature, tracking domain, prospects, duplicates, DNC, send window) and returns green/yellow/red — call it before every launch. `campaigns(action:analyze_body)` scores each step's copy for word count, CTA count, question count, link count, image count, and fabricated-claim patterns. `campaigns(action:analyze_subject)` scores a subject line for character count, spam-word hits, mobile truncation preview, personalization tokens, and surfaces the last 5 subjects from this account's campaigns. `prospects(action:readiness)` checks enrolled prospects for verified email, first name, company name, recent-contact history, and DNC membership before launch. `prospects(action:enrich_estimate)` shows the credit cost and missing-field breakdown before any enrichment spend is committed. `configure(action:signature_build)` is a structured signature builder: pass name/title/company/phone/website/linkedin and receive correctly formatted HTML (Gmail-safe black color, separator, Arial) + plain text ready to hand to `configure(action:signature_set)` — replaces ad-hoc HTML authoring. All six are live for every connected account on any paid plan; no plan or credit change.
- **v1.13 — 2026-08-20** — **Catalogued three capabilities that shipped since v1.12.** `saved_search` (define an ICP search once → the platform re-runs it daily/weekly and auto-enrolls new matches into a campaign, deduplicated, 1 credit per newly enriched contact) added to the Prospect-data list; `reports` (scheduled daily/weekly/monthly performance reports delivered by email) added to the Hygiene-&-account list; and the `notes` line now names the durable cross-surface **memory** actions (`remember`/`forget`/`list_memories`) that carry account facts and preferences between the MCP and the internal Senders chat. All three are live for connected accounts; no number, plan, or capacity change.
- **v1.12 — 2026-08-05** — **Aligned the send-path and hygiene copy to SP 7.1.** The system prompt now teaches `send_emails` as the single send chokepoint (campaigns for sequenced outreach; `send_emails` to launch, approve a manual batch, reply, or the occasional one-off) and no longer steers to `stage_send` — so this file stopped naming `stage_send` as "the" way to send (Single-plan line + tool catalog now name `campaigns`/`send_emails`). `stage_send` still exists as a plan-gated tool; it's just no longer the recommended path. Also scoped `analyze_email` honestly as a pre-send **hygiene** linter (does the copy *land*) and pointed the *does it convert* question at `strategy(play:"efficacy")`, matching the tool's own rescoped description. No number or plan change.
- **v1.11 — 2026-07-27** — **The data network's free half is now genuinely free.** `discover` was one tool carrying both the free actions (`catalog`, `quote`, `job_status`) and the paid one (`run`). MCP's read-only annotation is per-TOOL, not per-action, so the whole tool had to be marked as spending — which meant approval-gating clients raised a confirmation prompt just to BROWSE the catalog. Split into **`discover`** (read-only: `catalog`, `quote`, `job_status` — cannot charge, structurally) and **`discover_run`** (the single paid action). Also fixed: the v1.9 default-ON flip was **inert** — the three gates read the raw per-account override and ignored plan defaults, so every paying account without an override row still saw nothing while this file already said they had it. Gates are now plan-resolved and fail-closed. Updated the data-network paragraph and the tool-catalog lines.
- **v1.10 — 2026-07-27** — **Plan numbers are now generated, not hand-typed.** Price, capacity, and monthly credits for every plan are emitted from the platform's canonical plan data into a **single** marker-fenced region (`scripts/render_pricing_block.php`), so this file can no longer drift from what the platform enforces. What that closes: the **weekly ceiling** is now stated alongside the daily one (pacing a week off the daily number alone runs into a wall that was never mentioned here); the **monthly credit allotments** per plan are stated for the first time (the homepage had them; this file didn't); the **per-mailbox capacity rule is scoped to Single and Team**, with Business's own managed daily and weekly base stated separately (writing that rule as universal told a Business customer they were capped at a quarter of what they pay for); the **Free** row no longer reads "planning only" — free accounts genuinely run domain and spam checks and search the data network at no cost, and what they lack is sending and credits, which is now what it says; and the volume ramp is called an **automatic ramp** everywhere — the old "warmup" label collided with the separate paid warm-up network product. Two structural fixes came with it: the section formerly headed "What paid plans add" is now **"The plans — price, capacity, and credits"**, because the generated block covers the free tier too and Free was reading as a paid plan; and **"How capacity works" now points back at that one block** instead of carrying a byte-identical copy of it. Per-plan feature prose is unchanged; only the numbers moved into the generated block.
- **v1.9 — 2026-07-18** — **`discover` (the wider data network) is now on by default for every paid plan** (Single, Team, Business) — Benny flipped `feature_deepline_discover` from per-account dark-launch to default-ON for paid (commit `0b91547`). **Free accounts still don't get it** — they keep `find_prospects` + `acquire_prospects` (search-and-resolve) only. Updated the "The data network" `discover` paragraph, the tool-catalog `discover` line, and the honest-constraints line: dropped the "switched on per account / ask us to enable it / dark-launch" framing for paid, kept the honest "only offer `discover` when it's actually in the tool list" mechanic (a free account won't have it) and the never-name-the-operator rule. Paired with strategist KB v5.6.
- **v1.8 — 2026-07-17** — Added **The data network**, scoped into two honest tiers. *On for every connected account (all plans, incl. free):* `find_prospects` (search — free, no credits) and `acquire_prospects` (resolve/enrich — 1 credit per contact, waterfall resolution, empty lookups free, never re-charged). *The wider `discover` catalog (tech stacks, job/funding/growth signals, ad intel, phone, SMB/local, web research):* `catalog`/`quote` free, `run` spends credits (with a required `max_spend_credits` cap for sizeable or unpriced pulls) — a dark-launch feature switched on per account, with `find_prospects`+`acquire_prospects` as the always-on fallback. New "## The data network" section; `find_prospects`/`acquire_prospects`/`discover`/`inspect_data` added to the free-tier and tool-catalog lists; free tier reframed (network *search* is free, *resolving* spends credits); Single-plan line and honest-constraints updated (audience-dependent yield, sparse search fields, per-account `discover`, never-name-operator). Supersedes the v1.6/v1.7 datasets churn — the durable capability is the **network**, not the one-off dataset; curated datasets/promotions return with the redesigned model. Paired with strategy.md v1.3 and a strategist KB update.
- **v1.7 — 2026-07-06** — Removed **Datasets and promotions** (reverts v1.6): the YC-founders dataset was a one-off, pulled from prod, and the datasets/promotions model is under redesign — so the public surface no longer advertises a capability with no live data. Dropped the "## Datasets and promotions" section and the two tool-catalog lines (the `find_prospects`/`acquire_prospects` datasets framing and `configure(promo_redeem)`). Returns when the new promo model ships. Paired with removal PR `benny/remove-yc-founders-dataset` + strategist KB v5.2.
- **v1.6 — 2026-07-06** — Added **Datasets and promotions**: Senders-curated datasets via `find_prospects(datasets / dataset_search)` and `acquire_prospects(dataset_pull)` (priced per dataset, already-purchased data never re-charged, a promotion can cover a pull), plus `configure(promo_redeem)`. New "## Datasets and promotions" section + two tool-catalog lines. Shipped in the proprietary-data + promotions engine.
- **v1.5 — 2026-07-04** — Added **Automations** (`manage_automation`, Single+): trigger → action rules over prospect events (replies by sentiment, link clicks, unsubscribes, bounces, booked meetings) with ready-made recipes, a dry-run test, and Calendly/Cal.com booking intake via a per-account setup link. New "## Automations" section + tool-catalog line. Shipped in the WP2-40 Webhooks→Automations program.
- **v1.4 — 2026-07-01** — Added `domain_status` to the tool catalog (free-account and full-catalog lists): account-wide health of the user's own registered sending domains — DNS auth, blacklist, DMARC 7d/30d pass-rate, Google Postmaster spam-rate, and a per-domain score; no args for all domains, one domain for a live probe. Complements `check_domain` (survey the account vs inspect one domain). Shipped live in the 2026-06-30 domain-awareness deploy (WP-C); this line closes the llms.txt copy gap the 2026-07-01 /align pass found.
- **v1.3 — 2026-06-18** — Connected/OAuth mailboxes send 50/day each on every BYOM plan (was inconsistently described as 25 Single / 50 Team). Total BYOM capacity is now 50 x your mailbox count (Single up to 5 mailboxes = 250/day; Team includes 5 = 250/day). Runtime, copy, and KB now agree. See WP2-19.
- **v1.2 — 2026-06-13** — Industry targeting now leads with the `industries` parameter (Apollo tag IDs from `list_industries` / KB Appendix A) as the primary, reliable lever, matching live platform behavior; company-name/keyword matching is now described as a fallback. Noted that `funding_stage` and `market_segment` are not applied on prospect search.
- **v1.1 — 2026-06-06** — Product is referred to as **"Senders.ai"** throughout (was "Senders") so an assistant and its user resolve the name unambiguously. Corrected Team capacity to **250/day (50 per mailbox)**. Use-case playbooks now show a worked user↔AI back-and-forth per motion; added guidance to run motions collaboratively (legwork + draft from you, judgment + approval from the user).
- **v1.0 — 2026-06-05** — First public version: what Senders.ai is, the four plans, the tool catalog, the capacity model, honest constraints, and the no-account → free → paid hand-off.