Skip to content
Agentmart
Submit a tool

MELRA

Local MCP runtime for policy-governed, verified agent actions with durable receipts.

Connector by XAGI-Lab · GitHub account verified by the MCP Registry

IndexedApache-2.0Local (stdio)★ 12
Write a review

Chalk Labs

Need help deploying MELRA?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/xagi-lab-melra

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Local MCP runtime for policy-governed, verified agent actions with durable receipts.

Use cases: identity-access, agent-orchestration

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list, install config and code. Checked Oct 25, 2026.

Tools
11 read · 8 write · 5 delete · 1 not yet classified
Secrets
None asked for
Install scripts
None found
Runs as
Node.js package

Tools by what they do 25

Read from the source code; the running server may list different tools.

Read 11 · Look things up without changing anything.

  • browse
  • browser_read
  • list_files
  • melra_capabilities
  • melra_plan
  • melra_receipt
  • melra_task_status
  • melra_workflow_plan
  • melra_workflow_status
  • read_file
  • recall

Write 8 · Create, change, send or run things.

  • browser_click
  • browser_type
  • melra_task_cancel
  • melra_task_judge
  • melra_workflow_cancel
  • melra_workflow_control
  • move_file
  • write_file

Delete 5 · Delete, drop, revoke or wipe things.

  • approve
  • delete_file
  • melra_execute
  • melra_workflow_advance
  • run_command

Not yet classified 1 · We could not tell from the name and description.

  • remember

Network hosts it talks to 8

  • 169.254.169.254 · found in code
  • attacker.test · found in code
  • github.com · found in code
  • glama.ai · found in code
  • opencode.ai · found in code
  • raw.githubusercontent.com · found in code
  • records.invalid · found in code
  • xagilab.com · found in code

Files it touches

  • /dev/fd
  • /dev/null
  • /opt/homebrew/bin/tesseract
  • /proc/self/mountinfo
  • /proc/self/status
  • /tmp/evil.so
  • /usr/sbin/ioreg
  • /usr/sbin/screencapture
  • Your home directory (~)
  • C:\\Windows

Tools (25)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

  • approveApprove and run a held taskcan change or delete data

    Description is built at runtime.

  • browse

    Description is built at runtime.

  • browser_click

    Description is built at runtime.

  • browser_read

    Description is built at runtime.

  • browser_type

    Description is built at runtime.

  • delete_file

    Description is built at runtime.

  • list_files

    Description is built at runtime.

  • melra_capabilitiesMELRA capabilitiesread-only

    Report what this MELRA server can run and the policy in force. Call it once before melra_plan or melra_workflow_plan; it takes no input and changes nothing. Returns product, version and protocolVersion; tools; operations (the actions each operation kind supports); customEvidence (verifier names allowed in custom evidence); workflowNodes (node types for workflow definitions); and policy, including mode, workspaceRoot, unhinged, mutations, allowedCommands, deniedTraits, allowedDomains, circuitBreaker and credential names.

  • melra_executeExecute a planned MELRA taskcan change or delete data

    Description is built at runtime.

  • melra_planPlan a MELRA taskread-only

    Plan one bounded operation without running it: MELRA evaluates policy and stores the task in its own records only, so nothing outside MELRA changes until it runs. Send a goal, one operation and requiredEvidence, and leave constraints empty. Returns the task with its effect contract and status planned, awaiting_approval (approval holds the challenge) or policy_blocked (a normal result; policyDecision.reason says why); an invalid request, such as one naming an unregistered verifier or a file evidence path outside the workspace, fails the call instead. Keep its id: it is the taskId for melra_exec…

  • melra_receiptRead MELRA evidenceread-only

    Read a task's audit trail: returns { receipts, certificate? }. Pass taskId for all of that task's receipts plus its certificate, or receiptId for one receipt plus its task's certificate; receiptId wins if both are given. Each receipt records the capability, target, effect, principal, policy decision with any approval, timings, success, redacted observed effect and evidence; the certificate gives the overall result, such as VERIFIED_SUCCESS, PARTIAL or RECOVERY_REQUIRED. Use it to show what a task did; use melra_task_status for its current state.

  • melra_task_cancelCancel a MELRA task

    Cancel one task by taskId and return its record; nothing that already happened is undone. A planned or awaiting_approval task becomes cancelled and can no longer be executed. A task running in this MELRA process is sent an abort, which interrupts terminal run, browser, computer and http operations; the returned record predates it, so read the outcome with melra_task_status. Any other task, including one running in another process, is returned unchanged without error, and so is one another process starts at the same moment, which then runs: check the returned status. On a workflow step's task t…

  • melra_task_judgeJudge a MELRA task's semantic question

    Description is built at runtime.

  • melra_task_statusInspect MELRA task statusread-only

    Read one task's stored record by taskId (from melra_plan, or a workflow node's taskIds), in any status, policy_blocked and cancelled included, without changing or waiting on it. Returns the task: status, policyDecision, any approval challenge and approvedBy, the redacted request and result, error, attempts, receiptIds, certificateId, judgements (with judgedBy once answered) and timestamps. Use it to re-check a task after melra_task_cancel or from another session, or to re-read a pending challenge. Use melra_receipt for receipts and the certificate, and melra_workflow_status for a whole workflo…

  • melra_workflow_advanceAdvance a MELRA workflowcan change or delete data

    Run a workflow run's next wave; use it, not melra_execute, for workflow steps, calling again for later waves. It starts all ready nodes concurrently, each step under melra_execute's policy, approval and verification. Once the run has failed (a forward node failed, or was denied while no other's outcome is unknown), only compensations run, in reverse definition order, stopping at one that fails or is denied; otherwise nodes still waiting are offered again beside the newly ready ones, unless a waiting step's task was denied or otherwise moved since the last advance: then only the waiting nodes a…

Show 10 more tools
  • melra_workflow_cancelCancel a MELRA workflow

    End a workflow run by workflowId, the run id melra_workflow_plan returned. Its pending, ready and awaiting_approval nodes, compensations included, become cancelled and the run becomes cancelled; only once that is recorded are the tasks on those nodes, and any left planned or waiting under the run's steps, cancelled as melra_task_cancel would, so a cancel that fails changes no task, and one that fails after the run is cancelled reports it: cancel again to finish. Neither resume nor melra_workflow_advance continues it. It starts and undoes no effect: no compensation runs, and other nodes keep th…

  • melra_workflow_controlPause, resume, or suspend a MELRA workflow

    Halt a workflow run or lift the halt, without changing node state. action pause or suspend (same behaviour; only the status name differs) sets the run's status to paused or suspended unless it is verified_complete, cancelled or failed; melra_workflow_advance then fails with workflow_halted:<status> while the run is halted. Halt between advances: while an advance holds the run's lease, in this process or another, pause, suspend and resume fail with workflow_lease_held and change nothing, so retry once it returns. resume needs a paused or suspended run, sets the status its nodes imply and runs n…

  • melra_workflow_planPlan a MELRA workflowread-only

    Validate a multi-step workflow definition and create a new run without executing anything: MELRA writes only its own records and plans no tasks; nothing outside MELRA changes until the run is advanced. Planning fails on the graph errors it checks (workflow_* codes, such as a dependency cycle), a step melra_plan would fail on, a condition, until or delegation check no registered verifier can ask (evidence_verifier_unavailable:<name>) or on a file path outside the workspace (evidence_path_outside_workspace), or a step policy denies (workflow_policy_blocked:<nodeId>); steps needing approval ask d…

  • melra_workflow_statusInspect a MELRA workflowread-only

    Read a workflow run in any status by workflowId without advancing or changing it; workflowId is the run id melra_workflow_plan returned, the same one melra_workflow_advance, melra_workflow_control and melra_workflow_cancel take. Returns the run: status (such as planned, running, awaiting_approval, awaiting_input, paused, suspended, verified_complete, failed, recovery_required or cancelled), error, stateVersion, definitionId, definitionVersion, timestamps, and nodes keyed by node id, each with its status, taskIds and, where set, approval, prompt, input, iterations and error. Check it before mel…

  • move_file

    Description is built at runtime.

  • read_file

    Description is built at runtime.

  • recall

    Description is built at runtime.

  • remember

    Description is built at runtime.

  • run_command

    Description is built at runtime.

  • write_file

    Description is built at runtime.

Signals

Stars
12
Forks
6
Open issues
11
Weekly downloads
722
Last commit
Oct 24, 2026 (7 days ago)
Last release
Oct 16, 2026 (15 days ago)
Last package publish
Oct 24, 2026 (7 days ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Registry name
io.github.XAGI-Lab/melra
Version
0.3.0-alpha.14
Auth
No auth
Packages
npm @melra/cli
License
Apache-2.0
Repository
github.com/XAGI-Lab/melra
Website
xagilab.com/
First seen
Oct 24, 2026
Last checked
Oct 24, 2026
Source
Official MCP Registrygithub-graphql

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get MELRA running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying MELRA?

CallGet help
Cite this page

MELRA MCP server. Agent Mart. https://agentmart.dev/connectors/xagi-lab-melra (last updated 2026-10-24)