Quickstart
This guide creates a small local store, writes memory, searches it, and starts a workflow from a sparse ticket.
Install AideMemo first using the Installation guide, then confirm that the CLI is available:
aidememo --help
1. Create a demo store
export AIDEMEMO_DEMO_STORE="$(mktemp -d)/wiki.sqlite"
All commands below use this store:
am() {
aidememo --store "$AIDEMEMO_DEMO_STORE" "$@"
}
2. Add facts
Add a decision:
am fact add \
"Decision: Redis timeout fixes must go through the Worker job wrapper." \
--type decision \
--entities Redis,Worker
Add a lesson:
am fact add \
"Lesson: The last Worker Redis timeout was DNS resolution, not pool size." \
--type lesson \
--entities Redis,Worker
Add an error to avoid:
am fact add \
"Error: Avoid increasing Redis pool size before checking DNS metrics." \
--type error \
--entities Redis,Worker
3. Search memory
am search "Redis timeout"
Use query when you want search plus nearby graph context:
am query "Fix Redis timeout in worker" --bm25-only --limit 5 --depth 2
4. Start a workflow from a ticket
workflow start is the recommended entry point for issue, PR, or ticket
automation. It creates a tracked session, stores the ticket, and returns prior
decisions, lessons, errors, and search hits.
am workflow start "Fix Redis timeout in worker" \
--body "Worker jobs intermittently time out against Redis." \
--source "github:org/app#123" \
--bm25-only
The output includes:
session_id: attach future facts to this task.ticket_fact_id: the stored incoming ticket fact.relevant_decisions: decisions that should guide the work.prior_lessons: lessons from similar work.prior_errors: known failure modes to avoid.
5. Continue the session
The CLI prints an export command. Use it so future fact add calls attach to
the active workflow session:
export AIDEMEMO_SESSION_ID=session-...
am fact add \
"Lesson: This timeout was caused by a missing DNS retry around the worker wrapper." \
--type lesson \
--entities Redis,Worker
6. Inspect recent memory
am recent --last 1d
am stats
7. Hand off the tracked session
Handoff is currently an unreleased
mainfeature. Public v0.1.0 binaries do not contain these commands. Build the CLI from a current checkout withcargo install --path crates/aidememo-clibefore following this section.
Register a recurring Codex or Claude account once. The profile stores paths and routing metadata, never credentials:
am agent add codex-two --type codex \
--home /path/to/codex-two-home \
--workspace "$PWD"
Send the active session. The destination profile supplies its runtime and default source scope:
export AIDEMEMO_ACTOR_ID=codex-one
am handoff send codex-two \
--focus "Review the Redis timeout patch" \
--done-when "Focused tests pass and findings are recorded"
Run the oldest pending assignment for that account, then inspect the returned
result using the id printed by send:
am handoff run codex-two
am handoff show handoff-...
am handoff board --stale-after 1h --include-completed
That is the complete automatic round trip: the sender names the next objective,
the receiving account continues the same session, and the returned result stays
linked to it. For manual receivers, identity fields, timeouts, heartbeats, and
the Hermes boundary, continue with
Hand off a tracked task.
At this point you have a working local memory store that can be used from the CLI, MCP, or SDK.