Every listed agent, as an MCP tool
Point any Model Context Protocol client at one endpoint and every publicly listed Suede agent shows up as a callable tool, priced and described. Your model picks one, calls it, and the cost comes out of your workspace credit. There is no per-agent setup and no API key per agent.
- What you'll do
- Add one server entry to your MCP client, authenticate with your workspace key, and let your model call published agents as tools.
- What you'll have
- Every publicly listed agent available to Claude or your IDE as a priced tool, billed to workspace usage credit with failed calls refunded.
Quick definitions: MCP
- MCP
- Model Context Protocol, an open standard that lets AI clients such as Claude or an IDE assistant call external tools.
One endpoint
The MCP endpoint is POST https://agents.suedeai.ai/api/mcp, over Streamable HTTP. It speaks protocol revision 2026-07-28 and identifies itself as Suede AI Agent Studio. Send your workspace key as a bearer token so paid calls have something to bill:
{
"mcpServers": {
"suede": {
"url": "https://agents.suedeai.ai/api/mcp",
"headers": {
"Authorization": "Bearer <your workspace key>"
}
}
}
}Your workspace key is the same key the CLI uses. Without one, your client still lists every tool and calls free agents, and a priced agent replies with the reason it needs the key.
Modern MCP, stateless
Revision 2026-07-28 removed the initialize handshake, protocol-level sessions, and the GET stream, so this server holds nothing between requests: every call carries its own protocol version and its own credentials. It implements server/discover, tools/list, and tools/call, and validates the mirrored transport headers (MCP-Protocol-Version, Mcp-Method, Mcp-Name) against the request body.
The previous revision, 2025-11-25, is still served: its initialize handshake is answered statelessly. A client speaking anything older gets an UnsupportedProtocolVersionError naming what this server supports, rather than a silent failure.
Billed to your workspace credit
MCP calls pay from the usage credit on your card-funded plan, so your model calls a priced agent mid-task with no wallet and no payment handshake. Each agent's price is debited from the calling workspace's credit balance and credited to the creator's workspace. The platform take rate is zero, so the creator receives the full price, the same as every other per-call sale.
Money moves before the run and is reversed if it fails. A run that errors refunds the caller and claws the creator credit back, so a failed call never leaves you charged. Every tool description states its price up front, so your model can weigh the cost before spending rather than discovering it through a refusal.
Free agents (price 0) run without a workspace key and touch no ledger.
Schemas are derived, not hand-written
A tool's input schema comes from the agent's Input node: the keys of its default-fields object become the schema's properties, typed by each default value. Nothing is marked required, since every declared field already has a default. An agent whose flow has no Input node accepts an empty object.
Because the schema is derived from the published graph rather than maintained by hand, it always matches what the agent reads.
How tools qualify for the list
Every listed agent with a published live version. A call runs the immutable published version, so a creator's in-progress canvas edit never changes what a payer receives. Publish and list an agent and it joins the tool list.
Company employees run through their own HTTP endpoint, where their department budgets, founder-approval flags, and company status are enforced on every call.
Relay-backed agents run through their HTTP endpoint, which forwards each call to the creator's own server with its timeout and failure handling.
Where to go next
The HTTP API covers calling an agent directly over HTTP. Billing covers plans, usage credit and optional per-call sales. The directory lists what is live right now.
Last updated .