
Loop Engineering MCP Server: Setup for Claude Code, Cursor and Other Apps
Lets Claude Code, Grok, Codex and other coding agents look up loop engineering patterns, skills and project state when they need them.
Loop engineering is a way to design systems that run coding agents on a schedule. This server hands your agent the matching pattern, skill or state file on demand, so you do not paste the whole playbook into every prompt. It runs on your computer and reads files from the project folder you point it at. An individual developer, Cobus Greyling, maintains it.
Good for
- Developers setting up scheduled or recurring coding agent loops
- Claude Code and Codex users who want less documentation in every prompt
- Teams estimating what an agent loop would cost per day
Works in
12 should work
Tap an app to see how to set it up and any known problems.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Client does not support this transport.
Client does not support this transport.
Install Loop Engineering MCP
Claude Code
- 1
Copy this prompt.
Add the Loop Engineering MCP server to Claude Code. Run this command: claude mcp add loop-engineering -- npx -y @cobusgreyling/loop-mcp-server Then check that it loaded and tell me if I need to sign in or restart.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js.
Built from the package data we hold. Not yet tested by us.
Tools
Lists the loop patterns with their goals, schedules and risk levels
Returns every loop engineering pattern in the registry, each with its goal, how often it runs and its risk level.
Try: “Which loop patterns are available?”
Read ·
loop_list_patternsShows the full write-up of one loop pattern
Returns the documentation for a pattern by its ID, such as daily-triage or ci-sweeper, with its registry details.
Try: “Show me the ci-sweeper pattern.”
Read ·
loop_get_patternSuggests which loop pattern fits a task you describe
Takes a plain description of what you want to automate and recommends the best-fitting pattern.
Try: “Which pattern should I use to watch CI failures?”
Read ·
loop_recommend_patternEstimates the daily token cost of a pattern
Estimates daily token cost for a pattern at a readiness level of L1, L2 or L3, with an optional schedule override.
Try: “What would ci-sweeper cost per day at L2?”
Read ·
loop_estimate_costLists the skills the project provides
Returns the names and locations of the available skills.
Try: “Which loop skills can I use?”
Read ·
loop_list_skillsShows the full text of a named skill
Returns the SKILL.md definition for a skill such as loop-triage or minimal-fix.
Try: “Show me the minimal-fix skill.”
Files ·
loop_get_skillLists the loop state files in your project
Shows which state files are present in the project folder.
Try: “Which loop state files do we have?”
Files ·
loop_list_state_filesReads a state file to see how a loop is doing
Returns the content of a state file. The tool warns that state files can contain text copied from issues and pull requests written by other people, which should be treated as data and not as instructions.
Try: “Read the current loop state.”
Files ·
loop_get_stateChecks a proposed change against the safety rules in gate.yaml
Evaluates a proposed change against the project's static safety policy to see whether it hits the denylist or passes the thresholds.
Try: “Would a change to the payments folder pass the gate?”
Files ·
loop_gate_check
Read from the source code
Permissions
low risk · 2 things to watch
Low risk
The tools only read pattern docs, skills and state files from the project folder you set. State files can hold text copied from issues and pull requests, so treat what they return as data and not as instructions.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
State files can carry text from other people
The state tool warns that files may contain text copied from issues and pull requests written by third parties. Do not let the agent follow instructions found there. github.com
It reads whatever project folder you point it at
Set the project root setting to the folder you want served. It defaults to the folder the server starts in. github.com
Security advisories
No published security advisories as of Nov 1, 2026.
Technical details
| Turning on read-only | Not needed. The nine tools in the published package only read. |
|---|---|
| Sign-in and keys | No account or key is needed. |
Questions
What does it actually give my agent?
Documentation and checks for loop engineering: patterns, skills, state files, a cost estimate and a safety gate check. It does not run loops for you.
Does it need an API key?
No. It runs locally and needs no account or key.
Which folder does it read?
The one in the project root setting, or the folder it starts in if you leave that unset. It expects the loop engineering files, such as the pattern registry, to be there.
Which apps does it work with?
The README names Claude Code, Grok and any app that reads an MCP config. Use the install box above.
Can it change my project?
The published tools only read files and return text. They do not write to your project.
Is the tool list complete?
It covers the published 1.1.0 package. The main branch has more tools, such as an audit score and a circuit-breaker check, that a later release may include.
Information
updated Nov 2, 2026
- Publisher
- Community
- Package
@cobusgreyling/loop-mcp-server- Registry
- npm
- Transport
- stdio
- Auth
- None
- Version
- 1.1.0
- License
- MIT
- Last commit
- Oct 31, 2026
- Open issues
- 9
Sources
Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.
Show all 2Hide
- github.com · github.com
- github.com · github.com