Talkable MCP Server#
The Talkable MCP server lets you connect your own AI agent to Talkable over the Model Context Protocol and work with your program without logging into the dashboard. Point a client such as ChatGPT, Claude Code, Cursor, Claude, Codex, VS Code, Gemini, OpenClaw, Hermes Agent, Grok, Perplexity, or your own custom MCP client at the Talkable MCP endpoint and it gets every capability of Talkable AI: performance analysis and reports, customer support, campaign building and optimization, loyalty programs, Wallet passes, App Store integrations, and user management, or it hands a whole task to the Talkable AI Assistant and gets the result back.
The MCP server is customer-facing and account-scoped: an agent acts as you, sees only your account’s sites, and can do only what your Talkable role already allows in the dashboard. See permissions for the exact mapping.
Note
The MCP server exposes the same tools the Talkable AI Assistant uses. It is a channel — your permissions, not a separate product. Everything an agent does over MCP is scoped to your account and gated by your role.
What you can do#
Know your program — metrics with filters and segments, report exports, campaign and settings visibility including the settings history, purchase tracking, product questions with citations.
Resolve support requests — look up a referral, person, or coupon; approve or void a referral; unsubscribe, resubscribe, or anonymize a person; expire a coupon.
Build and optimize campaigns (Early access) — create campaigns, incentives, placements, copy, templates and emails, images, A/B tests end to end, coupon lists, test purchases, launch readiness, site settings.
Build Wallet passes (Early access) — pass design, notifications, publishing to installed passes, pass reporting.
Build loyalty programs (Early access) — program setup, earning rules, tiers, rewards and redemption, member experience, members, loyalty reporting.
Integrations and webhooks — App Store integrations installed, configured, activated, and tested; Custom App sends campaign events to your systems.
Manage users — invite users, set site roles, remove access, send password resets, manage account admins.
Work your way — run saved skills; delegate a task to the Talkable AI Assistant with
prompt_ai_assistantand collect the result.
The capabilities an agent sees depend on the scopes of the credential, your role on each site, and which early-access capabilities are enabled for the site. See permissions.
Things to try#
Once connected, ask your agent in plain language. For example:
For marketers
“What drove referral revenue growth in June versus May?”
“Clone the Spring Sale campaign as Summer Sale in test mode, with a $25 advocate reward.” (Early access)
For support
“Look up jane@example.com and show her referrals with their status.”
“Expire coupon SUMMER15-7GH2; the promotion has ended.”
For developers
“List my Talkable sites, then compare friend conversion for the last 30 days across them.”
“Install the Custom App, send referral approved events to https://hooks.example.com/talkable, and test it.”
How writes work over MCP#
A credential with mcp:read only never sees write tools. A credential with mcp:write
applies changes immediately, without a preview step, so grant it only to agents that need it. A task delegated with prompt_ai_assistant runs with
the credential’s permissions and applies its changes without waiting for approval. Every call is
logged with the credential, the tool, the time, the outcome, and the source.
See also
Talkable AI — what the assistant and your agent can do, and how it stays in your control.