Kotodian / siege

What is siege?

Siege is an AI-powered agent development tool built as a TypeScript web app that structures Claude Code or Codex CLI into a full dev lifecycle: describe plans, generate technical schemes, build Gantt schedules, execute tasks sequentially, review git diffs with inline AI findings, and auto-generate tests. It persists everything in SQLite with a visual dashboard for projects, folders, and progress—solving the chaos of one-off AI chats by adding continuity, breakdowns, and quality gates. Import plans from GitHub issues, Notion, Jira, or Markdown for seamless ai powered agentic workflow for project management.

Why is it gaining traction?

It stands out by wrapping raw AI agents into a clickable UI with Gantt charts, PR-style diff viewers, and one-click AI fixes—features missing from terminal-only tools like Claude Code. Multi-provider support (Claude, GPT, GLM) plus session reuse makes ai powered agents the future of computing feel practical today, especially for ai powered projects github. Early adopters love the import from GitHub or Jira turning issues into executable plans.

Who should use this?

Solo devs or small teams prototyping with AI agents, especially those hacking on ai powered github repos or managing agentic workflows manually. Backend engineers tired of context loss in AI chats, or full-stack folks wanting visual scheduling before code hits the repo—perfect if you're cloning GitHub projects and need structured execution without Jira overhead.

Verdict

Promising for ai-powered development but immature at 11 stars and 1.0% credibility—docs are solid with screenshots, but expect rough edges in edge cases. Try it if you're deep into Claude Code; skip for production until more battle-tested. Self-host with npm run dev for quick wins.