01
They open the agent panel in your product
Inside their own authenticated session.
MCP customer support for AI agents
Your users have agents. Serve them the same answers.
Your user connects their own Claude, OpenAI, Hermes, OpenClaw or custom agent to a scoped endpoint and gets your capabilities as typed tools.
Supported agents
Claude
Add Peeve as a custom MCP connector.
OpenAI
Connect over MCP; call your tools.
Hermes
Connect the Hermes agent over MCP.
OpenClaw
Connect the OpenClaw agent over MCP.
Any MCP client
Anything that speaks MCP connects and gets your tools.
One registry, two renderers
One capability registry
invite_teammate
change_plan
download_invoice
list_customers
Defined once, from the guided crawl and your config. Neither renderer is maintained separately.
Two renderers
The cursor
Renders it as motion: travel, pause, click, narrate.
The MCP server
Renders it as a typed tool with a dry-run effect string.
Agents get the same permissions as your users, never more.
How does a user connect their agent?
02
Peeve mints a personal endpoint
mcp.peeve.ai/northwind/u/8f2a
03
They paste it into their agent
Revocable by them, visible to you, rate-limited per grant.
Published automatically and regenerated by the same drift detection: llms.txt, AGENTS.md, and the MCP server card. Growth gives 5 grants per user; Scale is unlimited.
Why teams choose Peeve
Serve the users you can't see
Your users' agents get your capabilities as typed tools with stated effects, not a chat widget they have to scrape.
One registry, zero duplication
The same capabilities render as a cursor for people and typed tools for agents, defined once and never maintained twice.
Scoped and revocable
Each grant inherits exactly that user's permissions, is rate-limited per user, and can be revoked by them or seen by you.
14-day trial, no cardInstalls in an afternoonRead-only until you widen itKill switch in one commandCancel anytime, no lock-in