Connect your own system to AIML.chat agents by exposing it as an MCP server — your internal API, database, or service becomes a set of tools the agents can call, with a human approving every write. If a marketplace connector doesn't exist, build your own.
Why a custom MCP server
The Model Context Protocol is an open standard, so anything you can wrap in an MCP server becomes available to your agents: a pricing engine, an internal lookup, a provisioning action, a proprietary database. Read-only tools let agents answer with live, accurate data; mutating tools let them take real actions — always behind AIML.chat's write-confirm flow, so the agent proposes and a person approves before anything changes. Grounded in your content plus your own systems, the agent gives specific, cited answers instead of generic ones, and you keep full control over what it can see and do.
Connect it in three steps
- 1Build or host an MCP server that exposes your tools (the agent SDK includes a server template to start from), or use the Custom Webhook option in the marketplace.
- 2In AIML.chat, open MCP → Marketplace → Custom and add your server's URL plus auth (none, bearer token, or a custom header) — stored encrypted.
- 3Attach your tools to an agent and test. Mutating tools automatically require human approval.
Because every action is scoped by you and writes require approval, you expose exactly what you intend and nothing more. You start on a real free tier and scale as usage grows — see pricing for the agent and message limits, and the MCP docs for the server contract and how write-confirm works.
Do I have to write code for a custom MCP server?+
To expose custom tools, yes — but the agent SDK includes a server template, or you can use the Custom Webhook option for simpler cases.
How is my server's auth stored?+
Credentials are encrypted at rest and only decrypted in-process at call time; the agent never sees them.
Is there a free plan?+
Yes — start on the free tier. See pricing for higher agent and message limits.