Skip to main content

Daily workflow

The Engram daily loop is intentionally boring: load memory at the start, search when you need more, save when something durable emerges, and audit at the end.

Start of session

/engram load --for-agents "current task"

Or from the terminal:

engram load --for-agents "<task>"

The agent should reply with a compact count line such as Engram loaded: 8 memories / 24 total related memories. unless the human asks for IDs, rules, or raw output.

During work

Search when the task changes or you suspect project knowledge is missing:

/engram search "topic I might be missing"

Preview which memory files would route without printing their contents:

engram load --dry-run "<query>"

Return every visible routed match instead of the compact limit:

engram load --all "<query>"

Save one durable fact

/engram save knowledge

engram save captures the best single memory candidate, automatically updates a matching memory or creates a new one, and always shows the A/B/C approval gate before writing.

Save several memories from a session

/engram save-session
/engram ss

Provide candidates in this shape:

TYPE: rule | TEXT: Always run tests before release. | CONTEXT: Created from release planning so future agents preserve the test gate.
TYPE: knowledge | TEXT: Release notes live in CHANGELOG.md.
TYPE: workflow | TEXT: When releasing, run tests, update changelog, then tag.

CONTEXT: ... is optional. Add it only when it explains why the memory exists.

Mine recent chats

/engram save-session --query-level 3
/engram ss -a last 50 sessions

--query-level must be a positive integer. The agent may use up to that many recent human-agent chat sessions, including the current one, and must not invent unavailable history.

Accept-all shortcut

/engram ss -a

-a means the human explicitly approves every agent-recommended candidate. Agents must not add --accept-all unless the human requested it.

When an accept-all run reports related memories before writing, no file was saved yet. The agent should rerun with structured candidates:

TYPE: rule | TEXT: OAuth rotation follows release foundations. | DEPENDS_ON: release-foundation | LEVEL: advanced
TYPE: knowledge | TEXT: Invoice retries use exponential backoff. | UPDATE: invoice-retry-baseline

Role routing

Save role-specific memory:

engram save --role frontend ...
engram save-session --role backend ...

Tune role routing:

engram set-role frontend
engram set-role backend security
engram set-role

When engram set-role ... or engram set-rule-variant ... succeeds, the CLI returns an Agent action: line. Engram-aware slash adapters and MCP hosts should immediately rerun engram load "<current task/request>" and treat that result as replacing prior Engram-loaded context.

End of meaningful work

Check Engram health, report invalid memories, and propose anything worth saving from this session.

Useful commands:

engram upgrade
engram verify
engram repair
engram graph "<topic>"
engram quality-check
engram archive --reason "<why>" <id-or-file>

Next steps