# Hermes Hub

Hermes Hub is a control plane for hosted, multi-tenant Hermes Agent runtimes.
It provides REST management APIs, an MCP connector, A2A edge access, tenant
run orchestration, shared storage, skills, cron jobs, managed secrets, and
agent policy controls.

Use these documents for external client and agent integration. They are
sanitized for public documentation and intentionally omit host deployment
internals.

## Docs

- [External REST API Contract](https://hermes-hub-context7.pages.dev/docs/external-api.md): Client-facing REST endpoint groups, authentication, response conventions, and endpoint purposes.
- [MCP Contract](https://hermes-hub-context7.pages.dev/docs/mcp-contract.md): MCP endpoint, transport, authentication, tools, inputs, and safety notes.
- [Hermes Agent Tenant Environment](https://hermes-hub-context7.pages.dev/docs/agent-environment.md): What tenant agents can use inside the runtime, including workspace, company storage, LLM access, browser access, skills, and cron.
- [Flavor Authoring Contract](https://hermes-hub-context7.pages.dev/docs/flavor-authoring.md): Flavor manifest, config fragment, tenant assets, cron defaults, eval suites, and CLI workflow.

## Usage Guidance

- Prefer the contract docs above for LLM usage. The raw OpenAPI schema is for
  generated clients and may include operator-only fields.
- Do not infer host filesystem layout, host isolation mechanics, VM names, or
  credential storage locations from this documentation.
- Tenant runtime callback endpoints are not general client APIs.
- Flavor packages must not contain secrets, host-specific paths, or private deployment details.
- Never log or expose bearer tokens, runtime keys, provider keys, or secret material.