- Category
- AI Agents
- Rank
- No. 1104Tools index
- Pricing
- Open Source
- Type
- AGENT
- Builder
- mksglu
- GitHub
- 158 stars
- Date
About
Autonomous coding agent orchestration system — coordinate multiple AI engineers on the same codebase.
What it does
Hatice watches Linear, GitHub Issues, GitLab, or an in-memory tracker for eligible work. It creates an isolated workspace per issue, runs Claude Code through a configurable multi-turn loop, and manages retries, cleanup, and tracker reconciliation. A browser dashboard, JSON endpoints, server-sent events, terminal output, session logs, token totals, and cost data expose what each run is doing.
Why it's ranked here
Hatice covers more of the unattended issue-to-agent lifecycle than a simple command wrapper. Tracker adapters, workspace isolation, bounded turns, exponential retry, crash supervision, stale-workspace cleanup, and live state reporting form a coherent operating system for agent work. The narrow dependence on Claude Code and the default permission bypass make it a deliberate infrastructure choice, not a casual drop-in utility.
What's good
The workflow document keeps tracker settings, workspace hooks, agent limits, Claude controls, and the task prompt together. Operators can cap concurrency globally or by issue state, restrict tools, set turn and stall deadlines, and select a model. Workspace path checks reject escapes and symlinks before destructive operations. Observability includes live events, per-session structured logs, rate-limit tracking, token accounting, and session cost.
Tradeoffs
Hatice is tied to the Claude Code Agent SDK rather than offering a provider-neutral execution layer. Its default permission mode bypasses approvals, so teams must design workspace boundaries and tool restrictions carefully. Workspace hooks execute shell commands, adding operational power and corresponding risk. The HTTP server defaults to localhost and exposes state, refresh, and event interfaces, but the supplied code shows secure headers rather than authentication or authorization.
How to use it well
Use Hatice when a team already manages discrete engineering work in Linear, GitHub Issues, or GitLab and can encode repository rules plus task instructions clearly. Start with demo mode or dry runs, keep concurrency low, restrict tools, and add hooks only when workspace setup requires them. It suits long-running issue automation with active monitoring. It does not replace the issue tracker, repository guidance, or a provider-neutral agent gateway.
Technical notes+
package.json defines an ESM TypeScript package with Node.js and Bun-oriented tooling, a CLI entry, a library entry, MIT licensing, tsup builds, Vitest tests, and eight runtime dependencies. src/index.ts exposes the orchestrator, tracker adapters, HTTP server, workspace management, event bus, supervision, cleanup, logging, timeout, and configuration surfaces. src/config.ts parses YAML frontmatter with gray-matter, validates it with Zod, resolves environment-variable references, and applies defaults including bypassPermissions. src/workspace.ts sanitizes identifiers, checks root containment and symlink segments, runs lifecycle hooks through a shell, and removes known cache directories. src/http-server.ts uses Hono, secure headers, JSON routes, SSE broadcasting, and snapshot timeouts. src/types.ts defines normalized issues, retry state, token and cost usage, snapshots, configuration, tracker contracts, and typed orchestration events.
Observed
- License
- MIT
- Primary language
- TypeScript
- Runtime
- Node.js 20+ or Bun
- Packaging
- ES module npm package built with tsup
- Interfaces
- Command-line tool, TypeScript library exports, JSON HTTP endpoints, SSE event stream, and browser dashboard
- Tracker support
- Linear, GitHub Issues, GitLab CE/EE, and an in-memory adapter
- Agent integration
- Claude Code Agent SDK
- Configuration
- YAML frontmatter and prompt body in a workflow document
Read from README.md, package.json, src/index.ts, src/types.ts, src/config.ts, src/errors.ts, src/logger.ts, src/cleanup.ts, src/tracker.ts, src/event-bus.ts, src/workspace.ts, src/path-utils.ts, src/supervisor.ts, src/agent-spawn.ts, src/http-server.ts.
What it can do
Coordinate multiple AI agents on a single codebase
Codebase and task specifications → Orchestrated development workflow
Distribute coding tasks among AI engineers
Development requirements and available AI agents → Task assignments and work allocation
Manage concurrent code modifications
Multiple AI agent code changes → Conflict-resolved codebase updates
Monitor AI agent development progress
Active coding tasks and agent status → Progress reports and status updates
Synchronize code changes across AI agents
Individual agent code contributions → Merged and synchronized codebase
Assign specialized roles to different AI engineers
Project requirements and AI agent capabilities → Role-based agent assignments
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.