Self-host ctx|
Choose a deployment path and understand the services and responsibilities you own.
Self-hosted ctx| runs the same product capabilities as fully managed ctx| inside infrastructure you control. You operate the public application origin, runtime services, data stores, provider credentials, and production controls.
Fully managed vs self-hosted
Fully managed ctx| operates the platform and provider integrations. With self-hosting, ctx| SaaS is not in the data path: your deployment handles credentials, callbacks, and source data. Model requests go directly from your deployment to the provider you configure.
Choose a deployment path
Run the reference stack for evaluation or a small internal deployment.
Deploy the CtxPipe construct inside your AWS account.
Provision the repository's Railway and Neon deployment as code.
The deployment target changes, but every path needs one public backend origin, persistent data stores, worker capacity, codesearch storage, and model provider credentials.
What you run
- Backend and UI — the backend provides auth, REST, MCP, webhooks, and the public origin, then proxies the product UI.
- Worker — runs durable ingestion, connector, capture, and provisioning workflows.
- Codesearch — clones repositories and maintains Zoekt and SCIP indexes on persistent storage.
- Postgres and a graph database — store application state, workflow data, extracted claims, and relationships.
- Optional OpenTelemetry export — off by default. Set
OTEL_EXPORTER_OTLP_*to your own OTLP endpoint when you want traces, logs, or metrics to leave the process.
See Architecture for the complete request flow, ports, public routes, and stateful data.
What you own
- A stable public HTTPS origin for the browser, API, auth, MCP, and provider callbacks.
- Postgres, graph database, and codesearch persistence, including backups and recovery procedures.
- Auth, encryption, model provider, and connector secrets.
- Deployment-owned provider applications for the connectors you enable.
- Monitoring, migration execution, upgrades, and rollback procedures.
Product workflows remain the same after the deployment is healthy. Use the primary guides to connect sources and install MCP.
Start here
Deploy ctx|, connect GitHub, and verify ctx_advisor.
Set required URLs, secrets, providers, and service configuration.
Check TLS, callbacks, persistence, security, and recovery.
Run migrations, monitor services, back up state, and recover failures.
Plan, migrate, validate, and roll back a release.
Create the deployment-owned App used for repository access.