PairFoundry
PairFoundry is a collection of field-tested playbooks, prompt packs, and agent configuration files for developers who ship code with AI coding agents such as Claude Code, Codex, and other terminal-native agents. It provides a structured approach to setting up agent memory files like CLAUDE.md and AGENTS.md, creating skills and slash commands, and defining permission policies that fit the actual repository. The content is written from real projects and is versioned as tools evolve, with clear notes on where each approach stops working. The site offers three main packs: The Agent Operating Kit, Spec-to-Ship, and Review & Repair. The Agent Operating Kit covers everything that goes into the repo before prompting, including templates for four repo shapes and a skills library. Spec-to-Ship provides a nine-stage pipeline from a one-line idea to a reviewed pull request, with gate checklists to keep the agent on track. Review & Repair gives single-purpose review passes, a debugging protocol that prioritizes reproduction, and refactor prompts with blast-radius limits. Each pack is a downloadable folder of working files, designed to be used independently or together. PairFoundry also includes a free learning track and open documentation. It is built and maintained in the open, with a one-time purchase model for the packs and a 14-day no-questions refund policy. The focus is on solving the 'context gap' that causes agents to fail, not on replacing developer judgment.
- Category
- AI Agents
- Pricing
- Free
Problem solved
AI coding agents often fail because they lack the context of the repository, not because they can't write code. PairFoundry solves this by providing pre-built configuration files, prompt chains, and review passes that give agents the necessary memory files, skills, and workflow structure to produce accurate, maintainable changes.
Use cases
- Setting up CLAUDE.md and AGENTS.md memory files for a repository
- Creating custom skills and slash commands for Claude Code and Codex
- Running a nine-stage pipeline from spec to reviewed pull request
- Applying single-purpose review passes for correctness and security
- Debugging with a reproduce-first protocol to avoid wrong theorizing
- Refactoring code with explicit blast-radius limits
- Onboarding new developers to an agent-assisted workflow
Audience
- Developers
- Startups
- Enterprise
- Hobbyists
Pros
- Field-tested on real projects
- Versioned as AI tools change
- Explicit about limitations of each approach
- One-time purchase with 14-day refund
- Free learning track available
Cons
- Requires manual integration into repositories
- Packs are paid (not fully free)
- Focuses mainly on terminal-native agents like Claude Code and Codex
- Not a software tool; requires existing AI agent setup
Official website: https://pairfoundry.com/
Canonical LambdaBase page: https://www.lambdabase.com/tools/pairfoundry