Sentient-X
Sentient-X Docs

Sentient-X

Start here

Reach a useful, verified result without learning the whole platform first

Pick the door that matches your job today.

Authenticate in the browser, connect one explicit project, then work from the coding agent you already use.

sx auth login
sx project connect "$PROJECT_ID"

The CLI stores the revocable credential in the operating-system credential store. The workspace receives the project id and MCP configuration, not the key.

Start with this prompt:

Orient me to this Sentient-X project. Inspect the governed task, current experience,
latest Worlds evidence, autonomy artifacts, and Fleet state. Do not mutate anything.
Return the resource references, what is blocked, and the smallest safe next action.

Your first useful result

Name the outcome

State what a human or robot should be able to do, not which file should change.

Let the agent map the path

Ask for the owning pillar, canonical object, boundary, effect, and focused test. Inspect before mutating.

Change one vertical slice

Keep the task, lineage, safety behavior, and error semantics intact across the whole slice.

Demand evidence

Run the narrow test first, then the broadest relevant gate. A submitted operation is not a success until its terminal receipt and evidence say so.

Two commands worth remembering

make check      # repository-wide validation
sx watch REF    # reattach to durable platform work

Physical work stays explicit

An agent can prepare and inspect robot work. Hardware admission, motion authority, promotion, and safety decisions remain at their owning Fleet or Worlds boundary. Never convert a refusal into a fallback.

Next