Give your coding agent a MuleSoft engineering playbook¶

Mule Skills helps Codex, Claude, Copilot, Gemini, Cursor, and similar agents work with Mule 4 projects using the same lifecycle your team already knows: design, develop, test, package, review, operate, and troubleshoot.

Install in a Mule project See it in action

No agent replaces Anypoint Studio, Maven, MUnit, or your review process. The skills tell the agent how to inspect current-project evidence, choose the right tool, validate proportionately, and stop before a mutation that needs your approval.

MuleSoft developer¶

Describe the change in Mule terms. The agent routes between API design, Mule XML/DataWeave, documentation, MUnit, build, and review.

See a developer change

Test or release engineer¶

Use behavior-focused MUnit guidance, repeatable build gates, artifact evidence, and a separate approval boundary for Anypoint publishing or deployment.

Follow a workflow

Platform or support engineer¶

Analyze runtime health and incidents from authorized Anypoint evidence or supplied exports without turning missing telemetry into a confident conclusion.

Understand Anypoint access

The Mule lifecycle, with clear ownership¶

flowchart LR
    D[Design<br/>mule-api-design] --> I[Implement<br/>mule-development]
    I --> T[Test<br/>mule-testing]
    T --> B[Check and package<br/>mule-build + mule-lint]
    B --> R[Review<br/>mule-review]
    R --> A[Publish or deploy<br/>mule-ops + anypoint-connect]
    A --> O[Operate<br/>mule-ops]
    O --> X[Troubleshoot<br/>mule-troubleshooting]
    Docs[mule-docs] -. refreshes evidence .-> D
    Docs -.-> I
    Docs -.-> O

mule-build owns local checks, MUnit execution, packaging, local runtime work, and version/tag preparation. anypoint-connect, used through operational workflows, owns explicitly approved Exchange publishing and Anypoint runtime changes. See the complete ownership map.

Paste one instruction¶

Open your Mule repository in the agent and paste:

Fetch and follow https://raw.githubusercontent.com/Avinava/mule-skills/main/docs/agent-install.md
to install or update Mule Skills in this Mule repository. Detect the agent host and existing
configuration, preview the changes, preserve customized files, run the validation, and do not
commit or authenticate to Anypoint unless I approve it.

The agent inspects first, detects an existing install, previews file and MCP changes, avoids duplicate configuration, installs all eight skills, validates the result, and leaves commit and authentication decisions with you. Walk through the result.

What gets installed¶

Layer	Purpose	Credentials
Eight `mule-*` skills	Decide how to design, implement, test, document, build, review, operate, and diagnose	None
`mule-lint@2.0.0`	Mule standards, static analysis, XML formatting, RAML/OAS validation	None
`mule-build@3.0.0`	Readiness, MUnit, package, local runtime, version and tag preparation	None
`anypoint-connect@0.15.0`	Authorized Design Center, Exchange, Governance, telemetry, and lifecycle actions	Anypoint login only when needed

Use Node.js >=22.0.0 for the MCP servers; Node.js 24 LTS is recommended. The skills themselves are instructions and remain useful when an MCP server is unavailable—the missing tool becomes a visible validation gap.

Safety you can predict¶

Request	Default behavior
Review or diagnose	Read-only findings; no source or PR changes
Develop or repair tests	Implement only the requested scope and run focused plus proportionate checks
Build	Validate, run MUnit, and package; no version, tag, publish, or deploy
Release preparation	Preview/check first; version, commit, tag, or push only when explicitly requested
Publish, deploy, restart, scale, or rollback	Separate Anypoint scope, readiness probe, and explicit approval

Project facts stay in the Mule repository's AGENTS.md; reusable skill files stay neutral. Private hosts, tenant identifiers, credentials, payloads, and customer fingerprints do not belong in either the reusable bundle or examples.