Define the behavior, knowledge, Skills, and approved tools the remote endpoint should expose.
Hosted MCP
Hosted MCP server for an AI agent you own
A Prostir hosted MCP server is the published endpoint of an owned Agent, with its behavior, tools, knowledge, sessions, and access policy behind one HTTPS identity.
For the usual MCP path, model-token cost to the Agent creator is $0: the customer's AI client supplies the model under that customer's subscription and limits.
Early access
What is getting in the way
A local MCP setup adds repositories, environment files, tunnels, deployment, uptime, and credential work before the useful Agent can reach a real client.
What changes
Publish the Agent to its own HTTPS host and use its stable /mcp endpoint with OAuth-first access or an explicit scoped API-key alternative.
Hosted MCP
How Prostir gets you there
Build the Agent capability
Define the behavior, knowledge, Skills, and approved tools the remote endpoint should expose.
Choose the access policy
Configure exact OAuth email or domain grants, or issue a scoped API key for a supported server-to-server use.
Publish and connect
Publish the Agent, copy its HTTPS /mcp address, and verify it with the specific client and account policy you plan to use.
Early access
What you can verify
- Every published Agent exposes MCP on its own canonical HTTPS typed host.
- New Agents start OAuth-protected, with exact email or domain grants and explicit scoped API keys as an alternative.
- The published Agent remains the runtime owner for sessions, knowledge, tools, access, billing, and telemetry behind the endpoint.
A good fit if
- You need a remote MCP endpoint for an Agent without operating a local server and tunnel.
- You can test against the exact MCP client, account tier, connector policy, and transport you plan to support.
Not the right fit if
- You need a generic arbitrary MCP server unrelated to a Prostir Agent product.
- You need a promise that every client, account, connector, or review policy accepts the endpoint without qualification.
Questions before you decide
01Will it work in every MCP client automatically?
The endpoint uses the published MCP contract, but each client can impose its own account, connector, authentication, certificate, or review requirements. Test the exact destination before making a universal claim.
02Can I expose it anonymously?
Only through the separate explicit Anonymous choice for an eligible free Agent after risk confirmation. Private OAuth access is the default.
03Who pays for model tokens, and what do they cost the Agent creator?
For the usual MCP path, model-token cost to the Agent creator is $0. When a customer uses the Agent through a supported MCP connector in ChatGPT or Claude, the model runs in that customer's AI environment under the customer's subscription and limits. The creator pays for the Prostir plan, not those model tokens. AI Credits are needed only when the creator separately chooses a Prostir-hosted model run, such as a supported channel or widget. BYOK is optional and bills the creator's own provider account. Paid external tools remain separate.
04How do I give a company or client access?
Share access to the published Agent, not a .claude folder or GitHub repository. Every published Agent has https://{slug}.ai.prostir.build and /mcp; grant exact users or domains through OAuth, or issue a scoped expiring API key for an approved machine integration. You can revoke access without asking recipients to install or maintain source files.
Explore related options
Early access
Move the Agent endpoint out of the local tunnel
Tell us the Agent, tools, target clients, and required access model. We will help you define the hosted MCP path and its real compatibility checks.