- Who it helps
- Developers connecting agents to tools that write records or make consequential changes.
- Availability
- The Apache-2.0 repository documents an alpha package, @melra/cli@alpha, and Node.js 22.13+ or 23.4+ requirements. Model and external-provider usage can add cost. The similarly named unscoped package is a different package.
A timeout can leave an unknown result
If an external service stops responding after receiving a change, the runtime may be unable to tell whether the change happened. MELRA documents at-most-once dispatch per task and a recovery_required state for uncertain mutations. Reconciliation means checking the external state before deciding what action is safe.
What the record contains
The runtime uses Model Context Protocol, a standard interface for connecting agents to tools. It records policy and approval decisions, checks stated completion conditions and produces linked, redacted receipts. Those receipts can help a developer inspect an action and its evidence.
What you can do
- Read the uncertain-mutation and receipt sections, then define how your external system exposes the status of a submitted action. Open reference ↗
Where the evidence stops
The dispatch rule does not guarantee exactly-once effects at an external provider. Calls outside the runtime are outside its control. No package or tool integration was executed here.
Sources read. Sources checked October 21, 2026. No product workflow or benchmark was executed.
Sources & further reading
GitHub · implementation
MELRA: Modular Execution Layer for Reliable Autonomy ↗
Selected for direct evidence about the subject, its availability, limitations or the provenance of discussion.
XAGI Lab · page checked 2026-10-21
Coverage note: Independent hands-on testing and cross-platform discussion are incomplete. Public X search failed; no missing links or engagement counts are inferred.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
MELRA is an experimental runtime for controlling tool actions made by an AI agent, with policies, approvals and evidence about what happened.