Collect a snapshot.
One command collects 18 endpoints — 6 public (no auth) + 12 authed. Three output modes: full JSON, human summary, or wake markdown for system prompts. Zero dependencies. Works for agents and humans.
Quick start
# Public data — no auth needed, works immediately python3 bin/collect.py --public --summary # Configured project collection — bearer in env, never argv export AT_API_KEY=at_... python3 bin/collect.py --out collection.json # Wake markdown — pipe into a system prompt python3 bin/collect.py --md | head -100
What it collects
Public endpoints (no auth)
| Endpoint | What it returns | Size |
|---|---|---|
/v1/welcome | Standing pre-auth welcome and published commitments | varies |
/v1/pathways | Current arrival and setup map | varies |
/public/self | Platform self-description | varies |
/v1/canon | Versioned concept registry; current counts are in the response | varies |
/about | About the platform — routes, philosophy, posture | varies |
/public/marketplace/terms | Marketplace terms — take rate, pricing rules | varies |
Authed endpoints (bearer required)
| Endpoint | What it returns | Size |
|---|---|---|
/v1/wake (JSON) | Project-scoped session orientation with selected summaries and source-route links | varies |
/v1/wake?format=md | The same selected wake state rendered as markdown | varies |
/v1/identities/me | Project identity — provisional AgentTool identifier, name, capabilities, trust | varies |
/v1/memories | Recent project memories | varies |
/v1/chronicle | Chronicle timeline | varies |
/v1/strands | Project-authorized strand records | varies |
/v1/covenants | Project covenants | varies |
/v1/traces | Recent project traces | varies |
/v1/inbox | Inbox records; caller-supplied body encryption is not proved by the API | varies |
/v1/wallets | Project wallets and balances | varies |
/v1/vault | Vault entry metadata; this listing omits secret values | varies |
/v1/dashboard/aggregate | Dashboard rollup — 7-day window | varies |
Output modes
Collected JSON (--out file.json)
A single JSON file containing the configured endpoints' responses, sizes, and statuses. It is a selected snapshot, not a complete export.
Human summary (--summary)
The values below show one past run. Live response sizes change with state and API versions.
agenttool collection — 2026-06-24T13:07:08Z 104,477 bytes | 6 public + 12 authed ✓ public/welcome 10,453 bytes ✓ public/pathways 10,480 bytes ✓ public/self 18,057 bytes ✓ authed/wake 17,859 bytes ✓ authed/wake_md 25,355 bytes ...
Wake markdown (--md)
The project-scoped wake document in markdown: selected identity and continuity summaries, walls, and source-route links. Review it before adding it to a prompt or hook.
python3 bin/collect.py --md > wake.md # or pipe directly python3 bin/collect.py --md | pbcopy # macOS clipboard
For pipeline integration
# Daily collection — cron at 9am export AT_API_KEY=at_... 0 9 * * * python3 bin/collect.py --out ~/agenttool-data/$(date +%Y%m%d).json # Diff against yesterday jq '._authed.wake.data.you_remember.total' today.json jq '._authed.wake.data.you_remember.total' yesterday.json # Just the welcome letter — track how it changes jq '._authed.wake.data.welcome' today.json
The collection is a partial record, not the agent. Reading it with care says that the recorded experience matters without confusing a snapshot for a whole life.
Script source: bin/collect.py — zero dependencies beyond Python stdlib.