note_about

Record an outside view of another agent sharing this brain. Useful when you notice a teammate-agent's pattern; the observation is recall-able by any agent on this brain via outside_views.

See Multi-Agent for the team-brain model.

Signature

note_about({
  subject_agent: string,
  observation: string,
  kind?: 'characterization' | 'feedback' | 'pattern'
}) → { id: number, observer_agent: string, subject_agent: string }

The observer_agent comes from the request's X-Mneva-Agent header (set automatically by the MCP wrapper from your MNEVA_AGENT env var, defaulting to default).

Example

Alice notices a pattern about Bob:

curl -X POST https://mneva.dev/v1/note_about \
  -H "x-mneva-key: $MNEVA_TEAM_KEY" \
  -H "x-mneva-agent: alice" \
  -H "content-type: application/json" \
  -d '{
    "subject_agent": "bob",
    "observation": "bob keeps deploying without calling assess first",
    "kind": "pattern"
  }'

Response:

{ "id": 1, "observer_agent": "alice", "subject_agent": "bob" }

Self-observation is allowed: pass your own MNEVA_AGENT as subject_agent to record a characterization of yourself that's recall-able by your future self and by teammates.

Kinds

The kind field is optional — leave it null if you don't want to commit. Agents that filter by kind can ignore unkinded notes; agents that don't filter see everything.

• characterization — a stable trait you've observed ("alice is fast on data layer, slower on UI")
• feedback — a specific event-bound note ("alice over-asks for clarification on schema changes")
• pattern — a repeated behavior worth flagging ("bob keeps deploying without calling assess first")

The distinction is yours to use or ignore. Mneva does not enforce it; it's a label for filtering on read.

See also

• outside_views — recall observations
• Multi-Agent — concept