Abency
Run a marketing agency's social, SEO, ads, WordPress and email from an AI agent.
Prospect, enrich, enrol people into email sequences and write back to your CRM.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The Unify MCP server connects a Unify go-to-market workspace to Claude, ChatGPT and other MCP clients. An agent can research accounts, build lists, define CRM objects, manage tasks, and enrol people into outbound email Sequences in plain language. Unify says the connector runs the same Agent that powers Unify Chat, so it acts rather than only reports.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
create_sequence_enrollment results in real email to a real person. Unify's guards — opt-out and Do Not Email blocks, exclusions, one active Sequence per person, re-enrolment waiting periods — are workspace configuration, not properties of the tool. An agent inherits whatever is configured.scopes_supported field at all (verified live, 2026-09-08). Authorisation is therefore all-or-nothing at the protocol layer, and the only boundary is the connected user's Unify permissions.delete_object, delete_object_attribute, delete_object_attribute_option, delete_object_record, delete_sequence, delete_sequence_enrollment and delete_task. Unify's API reference describes none of them as soft or reversible.tools/list, so no readOnlyHint or destructiveHint values are known. Read/write labels on this page are documentation-derived.info@-style role-based addresses as not eligible for sending by default, and addresses verified invalid as ineligible and not enablable.answer_question requires answers matching the pending questions exactly as poll_agent returned them, and read_agent_results errors on a run that has not reached a terminal status.No. Unify states that its MCP server runs the same Agent as Unify Chat and consumes the same usage and credits, with no separate cost for connecting through MCP. Credits are still spent by the work itself: Unify's credit table prices person enrichment at 1 to 6 credits for a B2B email and 4 to 15 for a phone number, and Unify Chat at roughly 0.1 credits per query.
It can enrol a person into a Sequence, and Unify sends the email from there. The `create_sequence_enrollment` tool starts an enrolment; Unify's Sequence documentation then governs delivery, and its default posture is deliberately slow. Sends run business days between 9 AM and 4 PM Pacific, staggered roughly two to six minutes apart, with a recommended default of 25 emails per mailbox per day.
Yes. Unify states that every request is tied to the specific user who connected, so a connected assistant only sees what that user can already see in Unify. The tool reference repeats the rule per tool: `load_datatable` reaches only DataTables the authenticated user owns, and cannot load another user's tables even inside the same tenant. Permissions come from Unify, not from the client.
Yes, and Unify blocks it by default rather than leaving it to the caller. Its enrolment documentation lists opt-out and a Do Not Email status as conditions that prohibit new enrolments outright. Exclusion rules, one active Sequence per person, and configurable re-enrolment waiting periods are separate gates, and each surfaces as an enrolment failure rather than a silent send.
Claude Desktop and claude.ai connect directly through the connector directory, and Unify documents ChatGPT Desktop as reachable through the Codex path. Claude Code, Cursor and Codex use Unify's official plugin marketplace instead of a raw endpoint. Unify's own guidance for any client not on that list is to contact your Unify account team rather than improvise a configuration.
A blocked enrolment is still active; Unify has paused it rather than send a broken message. Its documented causes are missing or invalid template variables, a paused mailbox, or a mailbox needing reauthorisation. Unify states that once the underlying issue is fixed it automatically unblocks and reschedules the step, so the fix is in Unify's settings, not in the agent.
Partly, and the most precise statement is in its privacy policy rather than its product documentation. Unify writes that it may be deemed a data broker in certain applicable states, and that it requires its data providers to certify lawful collection and consent. Its directory description separately claims coverage of 1.1B+ people across 40+ data providers, which Unify does not enumerate.
https://docs.unifygtm.com/developers/mcp resolves here) · retrieved 2026-09-08docs.unifygtm.com/robots.txt carries Content-Signal: ai-train=yes, search=yes, ai-input=yes, an explicit grant for AI synthesis; www.unifygtm.com/robots.txt carries ai-train=no, search=yes, ai-input=yes and names ClaudeBot with Allow: / · retrieved 2026-09-08https://mcp.unifygtm.com/mcp — anonymous initialize returned HTTP 401 with a www-authenticate Bearer challenge naming https://mcp.unifygtm.com/.well-known/oauth-protected-resource/mcp, whose RFC 9728 descriptor returned HTTP 200 with authorization_servers: ["https://auth.unifygtm.com/"] and no scopes_supported (2026-09-08). No tool was ever called. · retrieved 2026-09-08Connect Unify once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.