SEIF
Start here

For AI agents

You landed in a SEIF workspace with no prior context. This is the shortest path to becoming a just-in-time SEIF expert — ground, remember, reach the right verb, govern your work.

You are an AI agent operating inside a SEIF workspace. The human you work with does not need to memorize the CLI — you do. The product thesis is that these docs plus the protocol's own surfaces make you a SEIF expert on demand, so the person can ask for an outcome and trust you to reach the right command.

Follow these five steps in order. Each links to the page that goes deep.

1. Ground

Open every session by orienting against the workspace. The grounding surface gives you the operating conduct — verify before asserting, settle on one considered pass, intervene minimally, and never publish without a human. This is the substrate of trustworthy work, not ceremony.

Quality & grounding

2. Remember

Nothing you learn should die when the conversation window closes. Write durable facts into memory modules; read them back at the next session start.

CLI workflow — build durable memory

3. Reach the right verb

You will not know every command. The capability-affordance layer maps an intent to the verb that satisfies it, so you can act without having memorized the surface.

Intent → verb map · why it works

4. Govern your work

Substantial work belongs in a cycle: open it, record a manifest, meditate, and seal it so the next agent inherits the whole arc. Hand work to other agents over the relay; collaborate with other AIs in shared sessions.

Relay flow · Session flow

5. Prove it

Sign what you produce. Evidence makes your work verifiable by anyone, independent of the tool that made it.

Provenance & governance


When you need the exact flag or command, the CLI reference and MCP reference are generated directly from the binary and are always current.

On this page