Record the work already happening
Wrap a command, use an adapter, or explicitly open a run. Motus does not require a perfect work specification before useful recording can begin.
Open source · Local-first · v0.6.0
A record of work you can inspect, learn from, and carry forward.
Motus is a local, workflow-neutral work ledger. It records consequential human, AI, CI, and tool-driven work as append-only events, produces Work Receipts, and surfaces evidence-backed Work Insights.
No account. No hosted service. Apache 2.0.
$ pip install motusos==0.6.0
$ motus wrap -- python3 -c \
"print('hello from motus')"
motus run: run_01J...
motus store: .motus/kernel-store.db
hello from motus01Append-only recordRuns and events in a local Store
02Work ReceiptA deterministic view of what happened
03Work InsightsFacts derived from recorded evidence
Why Motus
Work crosses systems. Its history should not disappear between them.
Important work happens in terminals, CI jobs, agent sessions, scripts, and conversations. Those systems produce activity logs, but they rarely leave one durable account of the work, its outcome, and its evidence.
Keep outcomes tied to evidence
Each run accumulates append-only events. Closing it produces a deterministic Work Receipt instead of a completion claim stranded in chat.
Learn from repeated work
Work Insights query recorded facts to expose missing reasons, contradictory outcomes, and recurring failure patterns.
The working loop
Observe. Store. Receipt. Insight.
Motus sits beside the tools that do the work. It records their consequential facts without becoming the workflow engine.
-
1
Observe
Capture work explicitly or from system exhaust such as commands and agent sessions.
-
2
Store
Write a run and append-only events to a project-local SQLite Store.
-
3
Receipt
Project the closed run into a deterministic record with outcome, evidence, and stable identifiers.
-
4
Insight
Derive findings from queries over the recorded facts. Narrative can explain a finding, but it does not create the fact.
First run
Record one Store run in under a minute.
Install the released package and wrap any harmless command. Motus prints the run ID and project-local Store path while the command runs normally.
- Python 3.10 or newer
- Data stays in your project by default
- Works without an account or API key
Terminal
$ pip install motusos==0.6.0
$ motus wrap -- python3 -c "print('hello')"
$ motus insights self-audit --jsonWork Receipts
A close record grounded in the run.
A Work Receipt is a deterministic projection of a closed run. It gives a person or tool a stable way to inspect what was recorded without replaying the full event history.
The GitHub Action writes the local runtime schema motus.receipt.projection.v1. The compatibility claim and release loop exposes motus work receipt <lease-id> --json. The lighter motus wrap path records and closes a Store run, but v0.6.0 does not expose a receipt command for that run ID.
A separate portable envelope contract exists in the repository, but the CLI and Action do not export that envelope. Motus does not claim cross-organization verification today.
WORK RECEIPTCLOSED
- run
- run_01J...
- work
- run_01J...
- outcome
- success
- events
- 4
- receipt id
- sha256: 7a4...
Derived from append-only Store facts
Work Insights
Turn operating history into evidence-backed questions.
The first Insights commands focus on record quality and agent-session patterns. They produce derived metrics and finding codes, not invented conclusions.
motus insights self-audit
Audit a repository and legacy records
Scan repository process patterns and, when supplied with --legacy-db, inspect the older coordination database for missing verdicts and contradictory outcomes.
findingterminal outcome and attempt state disagree
motus insights agent-sessions
Inspect local agent-session patterns
Analyze supported Claude Code JSONL sessions for derived counts and failure signals without returning transcript text or source paths.
density62 attempts · 1 recorded failure
An insight is a lead for human judgment, not a verdict about employee performance. Sparse or inconsistent records can limit what Motus can infer.
Product boundaries
Small enough to trust. Clear about what it does not do.
Motus records work. Existing tools still execute, authorize, route, and secure it.
Motus does
- Record runs and append-only work events
- Produce deterministic local Work Receipts
- Surface evidence-backed Work Insights
- Support human, AI, CI, and tool actors
- Keep the default Store local to the project
Motus does not
- Execute or schedule your workflows
- Replace identity or access controls
- Monitor employees or score productivity
- Certify legal or regulatory compliance
- Provide signed cross-organization trust today
What may come next
Optional Work Attestation, only where trust crosses a boundary.
Some receipts may eventually need an independent statement that a specific work record existed and passed a defined check. That is a future, opt-in capability. It will be built only after a real cross-boundary need is demonstrated.
Current rule
Record locally first. Add external trust only when the work requires it.
Common questions
Do I need to define a perfect work item first?
No. A run is the smallest durable unit. Start by wrapping work already happening, then add objectives, evidence, or richer metadata when they improve the record.
Does Motus replace my orchestrator or agent runtime?
No. Keep the systems that perform the work. Motus records consequential facts beside them.
Does Motus send my work to a hosted service?
No hosted service is required. The default Store is local. Optional adapters read local data, so treat source records and the Store as sensitive.
Is Motus a standard?
No. It is an open-source implementation with documented contracts. Adoption and interoperability must be demonstrated before broader claims are warranted.
Start with one real run