{
  "markdown": "<h1 align=\"center\">Agentic SWE</h1>\n\n<p align=\"center\"><strong>Claude codes your PRs. You review the receipt. Then merge.</strong></p>\n\n<p align=\"center\">\n  <a href=\"https://github.com/agentic-swe/agentic-swe/actions/workflows/ci.yml\"><img src=\"https://github.com/agentic-swe/agentic-swe/actions/workflows/ci.yml/badge.svg?branch=main\" alt=\"CI\" /></a>\n  <a href=\"LICENSE\"><img src=\"https://img.shields.io/badge/License-MIT-blue.svg\" alt=\"License: MIT\" /></a>\n  <a href=\"https://nodejs.org/\"><img src=\"https://img.shields.io/badge/node-%3E%3D18-brightgreen.svg\" alt=\"Node\" /></a>\n  <a href=\"CHANGELOG.md\"><img src=\"https://img.shields.io/badge/version-3.3.0-orange.svg\" alt=\"Version\" /></a>\n  <!-- catalog-counts:start kind=badge-line -->\n  <a href=\"#subagents\"><img src=\"https://img.shields.io/badge/subagents-138%2B-purple.svg\" alt=\"Agents\" /></a>\n<!-- catalog-counts:end -->\n  <a href=\"https://agentic-swe.github.io/agentic-swe-site/\"><img src=\"https://img.shields.io/badge/docs-site-informational.svg\" alt=\"Docs site\" /></a>\n</p>\n\nAn open-source autonomous SWE pipeline that runs in your editor or CI, writes every decision into your repo, and gives you a shareable audit trail of what the AI did and why.\n\n**What you get:**\n\n- **Structured PRs** — Claude works through a state machine (lean / standard / rigorous), not one mega-prompt\n- **Cost-attributed decisions** — every phase has a dollar amount and an artifact in `.worklogs/<id>/`\n- **Audit trail** — `/receipt` renders a shareable summary suitable for a PR description, Slack, or a compliance ticket\n\n## Quickstart\n\n```bash\nnpm install -g @agentic-swe/agentic-swe\nclaude --plugin-dir \"$(agentic-swe path)\"\n```\n\nThen in Claude Code:\n\n```text\n/work Add retry logic to the API client\n# ...pipeline runs, opens a PR...\n/receipt\n```\n\n→ See [Install & first run](#install--first-run) for Cursor, Codex, OpenCode, Gemini CLI, and Claude Code plugin marketplace alternatives.\n\n## What `/receipt` looks like\n\n`/receipt` reads `.worklogs/<id>/` and renders the work item as markdown. Sample from `test/fixtures/receipt/lean-happy/`:\n\n```markdown\n# /work add-retry-logic — Add retry logic to the API client\n\n| Field | Value |\n|---|---|\n| Work ID | add-retry-logic |\n| Track | lean |\n| Status | completed |\n| Duration | 47 min |\n| Cost | $1.84 |\n| PR | https://github.com/example/repo/pull/142 |\n\n## Decisions made (6)\n\n1. **feasibility → lean-track-check** ($0.08) — lean signal → feasibility.md#L1-L20\n2. **lean-track-check → lean-track-implementation** ($0.04) — verdict: simple\n3. **lean-track-implementation → validation** ($1.33) — implementation complete → implementation.md\n4. **validation → pr-creation** ($0.21) — tests green\n5. **pr-creation → approval-wait** ($0.18) — PR opened\n6. **approval-wait → completed** ($0.00) — approved by suraj\n\n## Human gates respected (1)\n\n- `approval-wait` resolved by user at 2026-05-17T14:47:00Z — approved by suraj\n\n## Loop counters\n\n- `self_review_iter`: 0\n- `doubt_cycles`: 1\n- `code_review_iter`: 0\n\n## Verifiable references\n\n- All artifacts: `test/fixtures/receipt/lean-happy/`\n- Audit log: `test/fixtures/receipt/lean-happy/audit.log` (9 entries)\n```\n\nEvery line above is computed from `.worklogs/<id>/` — no LLM summary, no hallucinated PRs. Reproduce locally:\n\n```bash\nnode scripts/render-receipt.cjs --work-dir test/fixtures/receipt/lean-happy\n```\n\n**Docs:** [agentic-swe.github.io/agentic-swe-site](https://agentic-swe.github.io/agentic-swe-site/)\n\n---\n\n## Pipeline at a glance\n\nAfter **feasibility**, **`lean-track-check`** sets **`pipeline.track`** in **`state.json`**. Tracks merge into **PR creation** → **`approval-wait`** → **completed**.\n\n```mermaid\n%%{init: {'theme': 'dark', 'fontFamily': 'ui-sans-serif, system-ui, -apple-system, Segoe UI, sans-serif'}}%%\nflowchart TD\n    start([\"/work — start or resume\"])\n    feasibility[\"feasibility\"]\n    check{\"lean-track-check<br/>sets pipeline.track\"}\n    lean[\"Lean track<br/>lean-track-implementation<br/>validation · pr-creation\"]\n    std[\"Standard track<br/>design → verification → test-strategy<br/>implementation → self-review<br/>validation · pr-creation\"]\n    rig[\"Rigorous track<br/>design → design-review<br/>verification → test-strategy<br/>implementation → self-review<br/>code-review → permissions-check<br/>validation · pr-creation\"]\n    gate{{\"approval-wait<br/>human gate\"}}\n    done([\"completed\"])\n\n    start --> feasibility --> check\n    check -->|lean| lean\n    check -->|standard| std\n    check -->|rigorous| rig\n    lean --> gate\n    std --> gate\n    rig --> gate\n    gate --> done\n\n    classDef accent fill:#1f6feb,stroke:#58a6ff,color:#ffffff,stroke-width:2px\n    classDef step fill:#21262d,stroke:#30363d,color:#e6edf3,stroke-width:1px\n    classDef branch fill:#21262d,stroke:#388bfd,color:#c9d1d9,stroke-width:2px\n    classDef decide fill:#21262d,stroke:#d29922,color:#ffdfb8,stroke-width:2px\n    classDef gateNode fill:#21262d,stroke:#a371f7,color:#e6edf3,stroke-width:2px\n\n    class start,done accent\n    class feasibility step\n    class check decide\n    class lean,std,rig branch\n    class gate gateNode\n```\n\nCanonical transitions: **`state-machine.json`** and the fenced graph in **`CLAUDE.md`** (checked in CI).\n\n---\n\n## Install & first run\n\nBeyond the [Quickstart](#quickstart) above, alternate paths:\n\n**Claude Code (plugin marketplace)**\n\n```text\n/plugin marketplace add agentic-swe/agentic-swe\n/plugin install agentic-swe@agentic-swe-catalog\n```\n\n**Other hosts**\n\n| Host | How |\n|------|-----|\n| **Cursor** | `curl -fsSL https://raw.githubusercontent.com/agentic-swe/agentic-swe/main/scripts/install-cursor-plugin.sh \\| bash` |\n| **Codex** | [`.codex/INSTALL.md`](.codex/INSTALL.md) |\n| **OpenCode** | [`.opencode/`](.opencode/) |\n| **Gemini CLI** | `gemini-extension.json` · **`GEMINI.md`** |\n\nAfter enabling the plugin, run **`/install`** once to merge **`CLAUDE.md`** and an optional **`.gitignore`** entry for **`.worklogs/`**. Maintainers see [`docs/PUBLISHING.md`](docs/PUBLISHING.md).\n\n→ [Full installation guide](https://agentic-swe.github.io/agentic-swe-site/docs/installation) · [Golden path (~15 min)](https://agentic-swe.github.io/agentic-swe-site/docs/golden-path)\n\n---\n\n## Commands\n\n| Command | Role |\n|---------|------|\n| `/work` | Start or resume a work item |\n| `/plan-only` | Feasibility / design without implementation |\n| `/brainstorm` | Design-first exploration (optional UI server) |\n| `/write-plan` · `/execute-plan` | Plan bar then execution |\n| `/check budget` · `/check transition` · `/check artifacts` | Enforcement before phases / transitions |\n| `/subagent` | Browse / invoke specialists |\n| `/repo-scan` · `/test-runner` · `/lint` | Evidence helpers |\n\n**Full list:** [Usage](https://agentic-swe.github.io/agentic-swe-site/docs/usage) · **`commands/`**\n\n---\n\n## Subagents\n\n<!-- catalog-counts:start kind=total-line -->\n**138+ specialists** under **`agents/subagents/`**\n<!-- catalog-counts:end -->. **Auto-selected** from **`feasibility.md`** signals; manual **`/subagent invoke`** anytime.\n\n<!-- catalog-counts:start kind=inline -->\nAcross 10 categories — Language Specialists (29), Infrastructure (16), Specialized Domains (15), Quality & Security (14), Data & AI (13), Developer Experience (13), Business & Product (11), Core Development (10), Meta & Orchestration (10), Research & Analysis (7).\n<!-- catalog-counts:end -->\n\n**Details:** [Subagent catalog](https://agentic-swe.github.io/agentic-swe-site/docs/subagent-catalog) · [Catalog routing](https://agentic-swe.github.io/agentic-swe-site/docs/catalog-routing)\n\n---\n\n## Work state\n\n**`.worklogs/<id>/`** holds **`state.json`** (source of truth), **`progress.md`** (timeline), **`audit.log`** (append-only), and per-phase markdown (e.g. **`feasibility.md`**, **`implementation.md`**, **`validation-results.md`**, **`pr-link.txt`**).\n\n- **State over chat** — resume from files, not from thread memory alone.\n- **Evidence** — tie claims to commands, paths, or CI (`templates/evidence-standard.md`).\n- **CI parity** — **`scripts/work-engine.cjs`** enforces **`/check`**-style rules.\n\n---\n\n## Architecture\n\nA single **Hypervisor session** (this one) owns transitions, gates, and synthesis. Three **core agents** — **developer**, **git-operations**, **pr-manager** — carry bounded work. A **design panel** (architect, security, adversarial) reviews in parallel on the rigorous track. All consult the **135+ subagent** catalog, auto-selected from repo signals.\n\n→ [Architecture overview](https://agentic-swe.github.io/agentic-swe-site/docs/architecture) (full diagram)\n\n---\n\n## Extending · CI · License\n\n| Topic | Where |\n|-------|--------|\n| Extend pipeline | **`/author-pipeline`** · [`references/authoring-pipeline-capabilities.md`](references/authoring-pipeline-capabilities.md) |\n| CI | [`.github/workflows/ci.yml`](.github/workflows/ci.yml) — **`npm run ci`** locally |\n| Research basis | [`CLAUDE.md` — Research basis](CLAUDE.md#research-basis) |\n| License | [MIT](LICENSE) · [Licensing](https://agentic-swe.github.io/agentic-swe-site/docs/licensing) |\n",
  "bytes": 9032,
  "sha": "95f03ee4d34a788d42b447ba4b6687440de68156889ddf09d718a603815a8bc8",
  "repo_slug": "surajsfdc/agentic-swe",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_surajsfdc_agentic_swe_agentic_swe_4aadd331/readme"
}