Stop Making AI Agents Grind Through Huge Codebases — I Built a Deterministic Wiki Build System for Them

Description and use

repowiki is a deterministic CLI tool that orchestrates wiki generation for large repositories. It enables safe, parallel task distribution among multiple agents, ensuring documentation is stored as code, validated, and reviewable in CI.

Problem

Large codebases make AI agents struggle to generate reliable, up‑to‑date wikis: context doesn’t fit, interruptions reset progress, and parallel coordination is manual. Existing solutions either require cloud access, lock into a single IDE, or lack incremental, versioned output.

Outcome

repowiki enables reliable, parallel, and incremental wiki generation for large codebases without external API calls, ensuring documentation stays in sync with code and is reviewable via standard git workflows.

Workflow

• 1. Run repowiki plan to scan the repo and split it into per‑page tasks.
• 2. Execute repowiki next to atomically claim a task directory.
• 3. Use any LLM agent (Claude Code, Codex, OpenCode, or a human) to generate the page content within the claimed task.
• 4. Agent writes output to the task directory; repowiki validates anchors, line numbers, and citations.
• 5. If validation passes, repowiki finalizes the page; otherwise the agent rewrites the page.
• 6. Repeat until all tasks are completed.
• 7. Run repowiki assemble to build the final wiki site and commit it to the repository.

Evidence

repowiki is "a build system that generates a structured wiki for any repository."

Source

The task catalog, claims, and heartbeats all live on disk under <repo>/.repowiki/

Source

The output is wiki‑as‑code. Markdown + mermaid diagrams + file:// source citations with line ranges, committed straight into your repository

Source