janaraj/tnl

Overview

A TypeScript tool that lets AI coding agents propose structured English contracts for their behavior. You approve or reject each contract, it is saved to disk, and every future session reads the same rules. This creates a persistent, auditable agreement between you and the agent.

Best for

Best for
Developers who want consistent agent behavior across many sessions without writing code

Use cases

• Define and enforce coding conventions across agent sessions
• Audit what rules an agent has followed in past work
• Onboard new agents to a project's agreed constraints

How to use

Install

npx -y typed-nl <command>

Tools exposed

• get_impacted_tnls
• retrieve_tnl
• propose_tnl_diff
• approve_tnl_diff
• non-goals

Tested with

Claude Code, Cursor

Notes

A TypeScript tool that lets AI coding agents propose structured English contracts for their behavior. You approve or reject each contract, it is saved to disk, and every future session reads the same rules. This creates a persistent, auditable agreement between you and the agent.

7 stars on GitHub. Last updated 2026-04-26. Licensed MIT.

Use cases

• Define and enforce coding conventions across agent sessions
• Audit what rules an agent has followed in past work
• Onboard new agents to a project’s agreed constraints

Pros

• Persistent rules reduce repetitive agent corrections
• Approval workflow gives you control without micromanaging
• Simple structured English is easy to read and edit

Cons

• Requires manual approval for each new contract
• Only works with agents that support the contract format
• No built-in conflict resolution if contracts contradict

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.