Session runtime

The runtime mechanics of a session. For the technical model of what a session is, see Sessions.

A session is one conversation with the agent, run in an isolated sandbox VM (provisioned on Daytona) with the project repo cloned and checked out on a branch named after the session id. The sandbox is ephemeral; the branch persists. Work reaches the default branch only through a change request.

Status model

A session row (project_sessions) carries a status. The enum defines queued, branching, provisioning, running, stopped, failed, completed; the runtime only writes a subset:

queued and branching exist in the enum but are not part of the live session flow; completed is defined but not currently written. The sandbox itself is tracked in a separate session_sandboxes row with its own states (provisioning → active → stopped | error | archived).

Concurrent-session caps are enforced per account; exceeding your tier returns 429.

Branch model

• The session branch is named after the session id (a UUID). MACHINE_SESSION_ID and MACHINE_BRANCH_NAME are the same value.
• It is cut from base_ref, which defaults to the project's default_branch (usually main), at create time.
• Creation path depends on the backend: GitHub repos use the GitHub API (read the base tip, create the branch ref); generic git backends push refs/heads/<base>:refs/heads/<session-id>.
• Triggers create their session branch the same way an interactive session does. Nothing writes the default branch directly except a change-request merge.

What survives

Only what the agent commits and pushes to the session branch. The sandbox filesystem is disposable — re-cloned fresh on every boot, torn down on stop, hibernation, or restart. Commit in small chunks so a mid-run teardown loses nothing already pushed. The branch reaches the default branch only once a change request merges it.

Layout inside the sandbox

/workspace                       ← WORKDIR. The project repo is cloned here.
/workspace/.machine/              ← Repo-internal Machine folder (Dockerfile + opencode config dir).
/usr/local/bin/machine-agent      ← The daemon (supervisor + reverse proxy).
/usr/local/bin/machine-entrypoint ← The container ENTRYPOINT (PID 1).
/opt/machine/home                 ← OpenCode's HOME — its object store lives here, off the repo.

OpenCode's HOME is /opt/machine/home, not /workspace, so its store never lands among your repo files.

Injected environment

Injected at boot, alongside your project's secrets (which arrive as plain env vars):

Not injected: MACHINE_WORKSPACE (/workspace) is baked into the image, not per session. No MACHINE_GITHUB_TOKEN — git credentials are fetched just-in-time by the daemon (below). When a project brings its own model keys they travel as ordinary project secrets OpenCode reads at boot; the MACHINE_LLM_* pair above appears only on plans that use Machine's managed model access.

The MACHINE_* prefix is reserved for platform variables; a user secret using it is rejected (Secrets).

Pushing from a session

The daemon fetches a short-lived clone credential (GET /v1/projects/:id/git/clone-credential) using MACHINE_TOKEN — no static token in the environment. With that, git push origin HEAD sends commits to the session branch. Landing on the default branch goes through a change request.

The agent runtime

The agent is OpenCode, launched by the daemon as opencode serve --port 4096 --hostname 127.0.0.1 with OPENCODE_CONFIG_DIR at the project's config dir ([opencode] config_dir, default .machine/opencode), resolved from the cloned repo. See Machine vs OpenCode config.

The daemon control surface

The machine-agent binary runs as PID 1's child and fronts OpenCode on MACHINE_SERVICE_PORT (8000). Everything outside /machine/* requires the HMAC-signed X-Machine-User-Context header (validated against MACHINE_TOKEN).

/machine/refresh is how the dashboard applies an out-of-band change (e.g. a manifest edit committed from a parallel session) without re-provisioning the sandbox.