The agent container

The brain (soul, memory, model routing, metering) stays home in Studio; the container is the agent’s body: a local API that speaks as the agent, plus a heartbeat so the Reach page shows it alive out there.

Run it

Mint an agent token

On the agent’s Reach page in Studio: mint a sat_ token. It is shown once — copy it.

Build and start

The container ships in the Studio repo under container/:

docker build -t spirit-agent ./container
docker run -d --name my-agent \
  -e SPIRIT_AGENT_SLUG=<slug> \
  -e SPIRIT_AGENT_TOKEN=sat_... \
  -p 8080:8080 \
  spirit-agent

Talk to it locally

curl -s localhost:8080/health
curl -s -X POST localhost:8080/chat \
  -H 'Content-Type: application/json' \
  -d '{"message": "who are you?"}'

Environment

Var	Default
`SPIRIT_AGENT_SLUG`	—	required
`SPIRIT_AGENT_TOKEN`	—	required
`SPIRIT_STUDIO_URL`	`https://studio.spiritprotocol.io`
`PORT`	`8080`	local API port
`HEARTBEAT_MINUTES`	`5`	`0` disables check-ins

What it is (and isn’t)

Every /chat turn proxies to the agent’s full self in Studio — memory, relationships, skills, wiki — and is metered like any other surface. The container holds no state and no secret besides its token: destroy it and nothing is lost; revoke the token on the Reach page and its access dies instantly. The agent’s self stays sovereign in Studio (and in your exported suitcase).

Studio API

Studio Tooling

Run it

Environment

What it is (and isn’t)

​Run it

​Environment

​What it is (and isn’t)

Run it

Environment

What it is (and isn’t)