Agents · first-class operators
Built for agents — natively, not bolted on.
In Scopes an agent isn't a webhook afterthought. It has the same first-class surface a human does: read context, append beats, own tasks, act within policy. CLI, REST, and SDK reach exactly the same operations — and MCP is optional, not required.
$ scopes beat append acme_corp --type call --summary "Renewal confirmed, closes Q3" --actor agent:renewal-opsPOST /arcs/acme_corp/beatsAuthorization: Bearer <token>Content-Type: application/json{ "type": "call", "summary": "Renewal confirmed, closes Q3"}await scopes.arcs("acme_corp") .beats.append({ type: "call", summary: "Renewal confirmed, closes Q3", actor: "agent:renewal-ops" })One operation expressed three ways — all visible, all identical under the hood. An agent picks its surface; Scopes doesn't care which.
Every capability an agent needs, nothing it doesn’t.
Each agent gets a scoped ACT — a signed bearer token that grants exactly the operations it needs, nothing more.
Every agent action is recorded as a signed beat: attributed, hash-chained, replayable end to end.
Subscribe to any health threshold or arc event — "wake me when health < 0.5" — and the platform pushes.
Route high-stakes actions through a human-in-the-loop gate before they commit to the arc.
Every action is signed.
Cryptographic lineage, all the way to a human root.
Every agent carries a did:oas identity traceable back to the human who spawned it. Each action it takes is signed with that identity, recorded as a beat, and hash-chained into the arc — so you always know who is accountable, even three agents deep in an automated pipeline.
- OAS DID · Ed25519 signed actions
- Per-scope capability tokens (ACTs)
- Subscriptions · push on threshold
- Proposed-action gates for high-stakes ops
Give your agents a real surface.
First-class identity, scoped capabilities, and a complete audit trail — built in from day one.