Skip to main content
These pages are for running Nasiko yourself. The open-source server is the default; an Enterprise license adds the self-managed extras marked on each page. On Nasiko Cloud, Nasiko runs the server for you. Both local paths run the same Docker Compose stack from a clone of the repository. nasiko up wraps Compose and registers the result with the CLI. A remote cluster is just a URL — see Install and connect. You need Docker Engine with the Compose V2 plugin: the docker compose command, not the legacy docker-compose v1 binary. The Compose file waits on service health checks, which v1 doesn’t support. From a clone of the nasiko repo:
The dashboard is http://localhost:8080. Sign in as admin / changeme. First start compiles the server image and takes several minutes.

What to change in .env

The template’s development values are enough to try Nasiko on your own machine, including AGENT_JWT_SECRET for routing coding agents. Replace them before anyone else can reach the server: OPENAI_API_KEY is only needed for features that call a model with the server’s default configuration, such as dashboard chat, the routing engine, and deployed agents. The dashboard loads without it. Coding agents routed through an LLM config use that config’s stored key. Keep the S3_* values matched to the bundled rustfs service. They’re development credentials, not generated secrets. After editing .env, run docker compose up -d again. docker compose restart doesn’t reload the environment. Compose sets TEMPO_URL, LOKI_URL, CODING_AGENT_OTLP_ENDPOINT, MCP_GATEWAY_PUBLIC_URL, and LLM_GATEWAY_BASE_URL for you, along with the database, Redis, and S3 connection URLs. These override anything in .env. That is why coding-agent traces and model routing work on Compose without extra exports. Harnesses and the CLI on your machine connect to http://localhost:8080 (or the URL others use to reach the server), not the Compose-internal http://server:8080.

host.docker.internal on Linux

Docker Desktop on macOS and Windows resolves host.docker.internal, so deployed agents reach the MCP gateway with no extra setup. Docker Engine on Linux doesn’t. If a deployed agent logs [Errno -2] Name or service not known, recreate it with --add-host host.docker.internal:host-gateway, or set MCP_GATEWAY_PUBLIC_URL to the Docker bridge IP.

Services and ports

Network name: nasiko. Agent containers join it. For a Rust developer loop (infra in Docker, server via cargo): docker compose -f docker-compose.infra.yml up -d, then just run / just dev. Same ports; the server binds 8080 on the host.

nasiko up

Run these from the repository checkout. nasiko up creates .env from .env.example if it’s missing, runs docker compose up -d --build, waits up to two minutes for /health, and registers the server with the CLI as cluster local. nasiko down runs docker compose down, which keeps your data. To delete the data too, run docker compose down -v in the checkout. Configuration is the same .env the Compose path uses. See Install and connect if it doesn’t start.

Next

Configuration

Environment variables.

Backup and restore

What to snapshot.

Troubleshooting

Startup, port, and build problems.