Dialer, texting and email CRM for sales teams. The connector lets an AI app search leads, read their calls, texts and emails, add notes and tasks, and draft or send follow-ups, with a confirmation before anything is sent.
Scored across 25 tools
Most tools are distinct (get_pipeline vs get_figures vs list_tasks), but `fetch` explicitly retrieves 'a merchant's record, a call, a text thread, or an email' — the exact jobs of get_lead, get_call, get_texts and get_email. Similarly the bare `search` overlaps with search_leads, search_calls, search_texts and search_what_was_said. Descriptions do steer agents toward the specific tools, but the boundaries are genuinely blurred.
Consistent snake_case with meaningful verb prefixes (get_ single record, list_ collection, search_ query). The convention is readable and predictable. `fetch` and the bare `search` break the verb_noun pattern and duplicate get_/search_ semantics, keeping it from a 5.
25 tools is at the heavy end, but the domain is broad — leads, calls, texts, emails, lists, tasks, events, templates, waves, pipeline, figures, team. Each has a distinct resource, so the count is largely earned. It is still borderline large and could be trimmed by removing the redundant generic `fetch`/`search`.
Read coverage is excellent across essentially every CRM entity, but the surface is entirely read-only — no create/update/delete for leads, tasks, notes or deals. Worse, descriptions reference tools that are not present (resume_ai_run, preview_batch, preview_wave, run_wave), creating dead ends where an agent is told to call something that does not exist.
Read one result of search in full by its id (lead:…, call:…, texts:…, email:…): a merchant's record, a call with its transcript, a text thread, or an email.
Inputs: id (required).
Whether the AI caller is dialling in this workspace and on which list, how far its run has got, its calls a day, and today's calls, people reached, voicemails and callbacks. The run's id is what resume_ai_run takes.
Inputs: no named parameters.
One call in full: the lead and who was spoken to, who made it (a rep or the AI caller), when, how long, its outcome, the analysis (summary, facts heard, review notes, quotes), objections and how they went, the call notes, and the transcript with each turn's time and speaker (we are Dan to the merchant). Long transcripts come a window at a time: pass transcript_from to read on, or transcript: false to skip it. The recording plays in the app at the call's link; it is not given out here.
Inputs: call_id (required), transcript, transcript_from.
One email whole, by its id from list_lead_emails or search_emails (gmail:…, campaign:…, or an older imported id): sender, recipients, subject, file names and the full words. Read it before quoting or answering a message the list only showed the start of.
Inputs: email_id (required).
How the floor did over a period: dials, connects, conversations, talk time, texts sent, replies, emails sent, and booked, won or (with funding) applications and funded — per person, the AI caller and everyone, compared with the same days before. Whose rows follow the app: an owner, admin or reviewer sees everybody's, a manager their groups', a rep their own beside the team's average. A list narrows it to the leads on it. Use it for "how many dials did we make this week", "how am I doing against the team".
Inputs: range, compare, list_id.
Everything held on one merchant in one read: who they are and how to reach them (phones in the order they are dialled), status and owner, the workspace's own fields, where things stand and what they told us, call facts (what they want, how much, open positions, why they declined), the open task, the latest calls, notes, texts and email, bank statement figures and deals — each part only where this app may read it. Start here for any question about one merchant; follow the ids it names into get_call, get_texts, list_lead_emails or get_lead_timeline for more.
Inputs: lead_id (required).
What happened with one merchant, newest first, in one feed: calls (with outcome and summary), notes, texts and emails that came in, emails we sent, the task, deals. Use it for "what happened since" or to catch up on a lead; since is a day. Only what this app may read is in it.
Inputs: lead_id (required), limit, since.
Where the deals stand: each stage with its leads (days in the stage, days since the merchant last wrote, owner, next step), and with funding on, revenue from statements, lender answers, offers and what was funded this month and last. The leads are the ones this person sees; the stage counts and money are the team's, as on the Pipeline screen. Use it for "what's in the pipeline", "which deals are stuck".
Inputs: stage.
The text messages with a merchant, both directions, oldest to newest, with who sent each of ours (a rep, a batch, or the AI holding the conversation), pictures and files they sent named, whether the number said STOP, and whether the AI holds the thread. Give lead_id for the lead's latest thread (its other threads are named), or conversation_id for one thread; pass cursor to read older messages. A text the carrier took is not proof it arrived.
Inputs: limit, cursor, lead_id, conversation_id.
What is on the calendar, day by day in this person's zone: the calls booked with leads (when, with whom, whose lead, how it was booked, the meeting link, whether it happened) and the events on the workspace's connected Google Calendar (in full on the person's own calendar, as busy times on a colleague's). Also names the calls that are over with nobody saying how they went. Use it for "what's on today", "when am I meeting Acme".
Inputs: days, from.
The email exchange with one merchant, newest first: theirs and ours (from a rep's Gmail, from one of our addresses, or a campaign), with who sent each, the subject, file names, whether theirs still waits for an answer, and the first lines of each. Use get_email for one in full. Mail that reached no connected mailbox is not here, so a quiet exchange is not proof they never replied.
Inputs: lead_id (required), limit, cursor.
The leads on one saved list right now (a list is a rule, so who is on it changes as leads move), with how many there are in all, this person's view only. Use get_lead on any of them for the detail.
Inputs: list_id (required), limit, cursor.
The workspace's saved lead lists: id, name and their rules in words (who is on it, e.g. "Interested, Eastern time, not called in 7 days"). Lists are how the floor groups leads to call, text or work; read the leads on one with list_leads_in_list, or pass list_id to search_leads to search within it.
Inputs: no named parameters.
The notes people wrote on one lead, newest first, whole, with who wrote each and when. get_lead shows the latest five cut short; read this for all of them.
Inputs: lead_id (required), limit, cursor.
What needs doing: each lead's open task (what it waits on, the next action, the due day, a callback time) grouped as overdue, due today, waiting, or received (the merchant answered). Defaults to this person's own; whose=everyone shows every task on the leads they see. Use it for "what's on my plate", "who do I call back today".
Inputs: group, limit, whose.
The people in this workspace: id, name, role and time zone. The ids are what search_calls' by and other tools take for a teammate.
Inputs: no named parameters.
The workspace's text templates that are switched on, each with its id, its words as a merchant with no first name on file would read them, and its segments. A batch (preview_batch) sends one template, or two as an A/B.
Inputs: no named parameters.
The workspace's text waves: a list texted in touches on a schedule. Each with its id, whether it is on, who it texts, its touches, when it runs next and last, and what it has sent and heard back. Run one now with preview_wave, then run_wave.
Inputs: no named parameters.
Search the CRM for merchants, text threads and calls by a name, a phone number, an email address or what was said. Returns ids to pass to fetch.
Inputs: query (required).
Find calls, newest first: one lead's, a day's or a week's, by who made them (this person, a teammate, or the AI caller), by how far they got (reached a human, voicemail, no answer …), by length, or by words in the lead's name or the call's summary. Each row has the call id, lead, when, direction, length, outcome and a short summary; read one with get_call. For what was said across many calls, search_what_was_said is better.
Inputs: by, from, limit, query, +6 more.
Find emails across the leads this person sees, newest first: by words in the subject, the body, the sender's address or the lead's name; theirs or ours; only the ones still waiting for an answer ("who emailed us back and hasn't heard from us"); within days. Each row has the email id for get_email and the lead. For what was said across texts, calls and emails at once, search_what_was_said is better.
Inputs: from, limit, query, cursor, +4 more.
Find merchants (leads) by business name, contact name, email, phone number, or by what was said on their file ("asked about equipment" also finds "needs a new truck": words first, then meaning). Narrow by status, a saved list, starred, or only this person's own. Use it to get a lead id for get_lead; use search_what_was_said instead when the question is who said something rather than who someone is. Statuses are the workspace's own words: a status that does not exist is refused with the ones that do.
Inputs: mine, sort, limit, query, +4 more.
Find text messages across the threads this person sees, newest first: ours, theirs or both; by words in the text, the lead's name or a phone number; one lead's; within days. "Who did we text last" is direction ours with limit 1; "who texted us today" is direction theirs from today. Each row has the thread id for get_texts, the lead, when, who sent it (a teammate by name, or the AI) and the text. For what was said across texts, calls and emails at once, search_what_was_said is better.
Inputs: from, limit, query, cursor, +3 more.
Find merchants by what was said or written: one search over texts, call transcripts and summaries, emails, notes, what leads told us, and file reviews, first for the words (a quoted phrase, a minus to drop a word) and then for rows that mean the same thing ("asked about rates" also finds "what's the APR"). Each hit says which and where it is, with the line itself. Use it for "who mentioned a lease", "which merchants asked about rates", "did anyone say they'd call back Monday"; use search_leads for a name, number or email.
Inputs: query (required), from, kinds, limit, +2 more.
Who this connection works as: the person, their role, the workspace, what this app may read or change here, their time zone and which leads they see. Call it first when you need to know whose leads "my leads" are, or why a tool is missing.
Inputs: no named parameters.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Add one secure layer between your agents and this server.