Deployment Modes

Corrobore exposes one graph runtime through two process models: embed the engine inside a Rust application, or run it as a separate service. The HTTP API is the network contract of the service model, while the standalone executable is its supported operational entry point.

Choose a mode

Mode	Process boundary	Interface	Choose it when
Embedded Engine	Inside the host Rust process	corrobore_engine Rust API	A Rust application should own engine construction, policy, lifecycle, and direct calls without a network service.
HTTP Server	Separate server process	Authenticated HTTP/JSON API	You are integrating an agent, service, Python client, or other remote caller and need the route, authentication, limit, and response contracts.
Standalone Server	Separate corrobore process	HTTP/JSON plus operator CLI	You are deploying Corrobore as a durable service and need validated configuration, lifecycle commands, storage ownership, TLS, monitoring, backup, and upgrades.

All three paths share the same policy, budget, validation, query-planning, and execution layers. The choice changes who owns the process and how callers reach the runtime; it does not create a different Cypher implementation.

How HTTP and standalone relate

HTTP and standalone describe different layers of the same service deployment:

• the HTTP Server guide is the client and transport reference. Use it to understand routes, authentication, request limits, sessions, and response behavior;
• the Standalone Server guides are the operator reference. Use them to install, configure, start, supervise, persist, secure, back up, and upgrade the process that exposes that HTTP API.

The unified corrobore server start command is the supported standalone product entry point. The legacy corrobore-http-server binary remains a development compatibility target; it is not a second production deployment mode.

The Embedded Engine does not open an HTTP listener. The host application owns its lifecycle and calls the Rust facade directly. Choose the service model when you need language-neutral or remote access, centralized authentication, or operator-managed durability.

Detailed guides

• Embedded Engine — Rust facade, identity, policy, seed search, and deterministic export.
• HTTP Server — HTTP routes, authentication, limits, sessions, imports, exports, and error contracts.
• Standalone Server CLI — unified commands, lifecycle, security boundaries, and distribution entry point.
• Standalone Configuration — precedence, TOML fields, environment variables, CLI overrides, and defaults.
• Standalone Operations — native, Docker, systemd, observability, backup, restore, upgrade, and rollback runbooks.

For a first local run, continue with Getting Started. For container deployment, use Getting Started with Docker.