Context, scoped by design

Structured data, scoped and versioned by default.

Structured data is namespaced, versioned, and attached to deployments explicitly — read or read-write, checked every time it's touched.

Request access View docs

Deployment

↓

Knowledge scope

├─ namespace: support-docs

├─ namespace: tickets

└─ namespace: playbooks

↓

Runtime access

↓

Trace records usage

Context should not be a global bucket.

Production agents need real documentation, policies, tickets, and operational data. That access has to be scoped — by deployment, namespace, key, and permission level.

Unscoped context

→ unclear access, hard to audit

Scoped knowledge

→ explicit permissions, traceable usage

One primitive. Namespaced, versioned structured data.

Namespaces

Docs, playbooks, tickets, policies — organized by namespace and key, not a global bucket

Version history

Every write keeps its prior versions, so you can see exactly what an agent read or changed

Attach only the context a deployment is allowed to use.

Scope by deployment

Scope by namespace and key

Define read or read-write access

Narrow access further per run

Keep every access observable

knowledge configuration

structured_data:
  tickets:
    permission: read_write
    keys:
      - "ticket:*"
  playbooks:
    permission: read
  support-docs:
    permission: read

Knowledge access appears in the run timeline.

When an agent reads or writes structured data, sypho records it — alongside every skill, tool call, log, and result.

run.started

retrieval.call: tickets

retrieval.result: ticket:4821

tool.call: create_case

step.completed

run.completed

Context that's explicit, never accidental.

Explicit context

Attach the right knowledge to each deployment instead of relying on broad global access.

Namespace-aware scope

Limit access by namespace, key pattern, and permission level.

Versioned by default

Every write keeps its history, so nothing is silently overwritten.

Runtime control

Apply permissions to what agents can read or update.

Traceable usage

See knowledge queries and accessed context inside run traces.

Give agents the context they need — and the boundaries production requires.

Request access and help shape knowledge infrastructure for production agents.

Request access