Skip to content

PACE Framework

AI agents that plan, build, review, and ship your sprint — day by day.

What is PACE?

PACE (Plan · Architect · Code · Evaluate) is an AI-driven sprint execution framework. Instead of one monolithic AI generating an entire feature, PACE runs a structured team of six specialised agents that mirror the roles of a real engineering squad:

PRIME — Planning

Reads your sprint plan and generates a daily Story Card with user stories, acceptance criteria, and scope boundaries.

FORGE — Implementation

Writes production code using a tool-calling loop. Follows TDD, respects source directory boundaries, and produces a Handoff Note.

GATE — Quality

Runs your test suite, checks CI results, and evaluates every acceptance criterion. Issues SHIP, HOLD, or PARTIAL decisions.

SENTINEL — Security & SRE

Scans for hardcoded secrets, input validation gaps, missing timeouts, and unhandled error paths.

CONDUIT — DevOps

Reviews CI/CD workflows, Makefile targets, dependency lock files, and deployment configuration.

SCRIBE — Documentation

Keeps your context documents, README, and engineering docs in sync with what was actually built each day.

Why PACE?

Most AI coding assistants work in isolation — they write code but can’t run your tests, review CI, or check for leaked credentials. PACE closes that loop:

  • Structured handoffs: every agent reads the previous agent’s output, creating a verifiable chain of evidence.
  • Blocking gates: a failing test or exploitable vulnerability stops the day before code ships.
  • Advisory backlog: non-blocking findings accumulate and are force-cleared on dedicated review days.
  • Platform-agnostic: GitHub, GitLab, Bitbucket, Jenkins, or fully local.
  • LLM-agnostic: Claude, GPT-4o, Gemini, Bedrock, Ollama, Groq, and any LiteLLM-supported model.

Explore the Docs