Add MartinLoop to specialized tools

[Keesan12]

Adds MartinLoop as a specialized tool around AI coding agents.

MartinLoop is an open-source control and verification layer for coding-agent runs, with budget limits, verifier gates, failure triage, recovery evidence, and inspectable run receipts.

Project: https://github.com/Keesan12/martin-loop
MCP: npx -y @martinloop/mcp
License: Apache-2.0

[Keesan12]

Quick follow-up on scope:

The strongest fit here is as a specialized control / evidence tool around AI coding agents rather than as a base model or raw runtime.

The current public line is useful to people who already run coding agents and want to answer:

• what changed
• what failed
• how much spend happened before the first real code delta
• what evidence says the run should be trusted or stopped

That positioning has become sharper with the recent 0.4.x releases, especially the routing-economics and diagnostics work.

[Keesan12]

Refreshed the entry with current factual MCP/runtime details. MartinLoop is an open-source control and verification layer for coding-agent runs, with budget limits, verifier gates, failure triage, recovery evidence, and inspectable run receipts. The PR is still conflict-free and mergeable; would appreciate a review when convenient.