🎵 Risoluto

Turn your issue tracker into an autonomous coding pipeline.
Linear issues in → sandboxed AI agents out → PRs delivered.

_{Overview — active issues, recent events, and run statistics}

Board — Kanban view with workflow columns

Settings — Tracker, provider, sandbox, and Codex admin controls

Risoluto is a local orchestration engine that watches your project tracker for actionable issues, spins up sandboxed AI coding agents to work on each one, and delivers the results — all without you supervising a single prompt.

You create an issue  →  Risoluto picks it up  →  AI agent writes code  →  PR lands on GitHub

No cloud service. No SaaS. Your code stays on your machine. Risoluto runs on your local machine or a VDS you control, connecting your existing Linear project to AI-powered Codex agents running inside isolated Docker containers. It communicates with configured external services (Linear, model providers) via their APIs, but your source code and workspace data never leave your infrastructure.

flowchart LR
    A["🗂️ Linear / GitHub<br/><sub>Issue Tracker</sub>"] -->|"poll"| B["🎵 Risoluto<br/><sub>Orchestrator</sub>"]
    B -->|"create"| C["📁 Workspace<br/><sub>Per-Issue Isolation</sub>"]
    B -->|"launch"| D["🐳 Docker<br/><sub>Sandboxed Container</sub>"]
    D -->|"contains"| E["🤖 Codex<br/><sub>AI Agent</sub>"]
    E -->|"JSON-RPC"| B
    B -->|"persist"| F["💾 Archive<br/><sub>Attempt History</sub>"]
    B -->|"serve"| G["🖥️ Dashboard<br/><sub>Web UI + API</sub>"]
    B -->|"push"| H["🐙 GitHub<br/><sub>Branches + PRs</sub>"]
    B -->|"alert"| I["📬 Multi-Channel<br/><sub>Slack, Webhook, Desktop</sub>"]

    style A fill:#7c3aed,stroke:#6d28d9,color:#fff
    style B fill:#2563eb,stroke:#1d4ed8,color:#fff
    style C fill:#059669,stroke:#047857,color:#fff
    style D fill:#0ea5e9,stroke:#0284c7,color:#fff
    style E fill:#d97706,stroke:#b45309,color:#fff
    style F fill:#6366f1,stroke:#4f46e5,color:#fff
    style G fill:#dc2626,stroke:#b91c1c,color:#fff
    style H fill:#1f2937,stroke:#111827,color:#fff
    style I fill:#e11d48,stroke:#be123c,color:#fff

The fastest way to get running. No environment variables needed upfront — the web-based setup wizard handles everything.

1. Clone and launch

git clone https://github.com/OmerFarukOruc/risoluto.git
cd risoluto
docker compose up --build

2. Open the dashboard

Navigate to http://localhost:4000 — the setup wizard starts automatically.

3. Complete the wizard (3–5 min)

4. Create an issue and watch it run

Set a Linear issue to "In Progress". Within one poll cycle (~30s), Risoluto picks it up, launches a sandboxed agent, and shows progress on the dashboard.

pnpm install && pnpm run build && bash bin/build-sandbox.sh
node dist/cli/index.js --port 4000

export OPENAI_API_KEY="sk-..."   # API key path
# — or —
codex login                       # ChatGPT/Codex subscription path

Risoluto's active roadmap is tracked in #354 — Symphony v2 Feature Roadmap (2026 Research Reset). As of 2026-04-04, the reconciled local roadmap shows 71 open feature issues across 11 currently named bundle categories, excluding the active epic #354 and the long-range vision issue #62. The older #9 roadmap issue is preserved as historical context only.

The latest roadmap reconciliation closed already-implemented issues #254, #258, #260, #262, #275, #276, #278, #282, #286, #292, #299, #303, #307, #308, #315, #318, #319, #326, #333, #335, #346, and #375, while the remaining stronger partials were promoted from research to triage.

The newest still-open post-reset additions are #366 fanout/merge execution, #367 risoluto doctor (triage), #368 dependency graph UI, #369 per-step success criteria, and #373 the Docker Sandboxes executor backend (triage).

Risoluto runs in Docker with a zero-configuration start — or scales to a control/data plane split for advanced deployments.

docker compose up --build

All data persists in named Docker volumes:

For hot upgrades, multi-host workers, or scale-out scenarios:

flowchart TD
    subgraph Docker ["🐳 Docker Compose"]
        CP["🎵 Control Plane<br/><sub>:4000 — Dashboard, polling, config</sub>"]
        DP["⚙️ Data Plane<br/><sub>:9100 — Agent execution</sub>"]
    end

    CP -->|"HTTP + SSE<br/>Bearer auth"| DP
    DP -->|"Docker CLI"| CX1["🤖 Worker 1"]
    DP -->|"Docker CLI"| CX2["🤖 Worker 2"]

    style CP fill:#2563eb,stroke:#1d4ed8,color:#fff
    style DP fill:#059669,stroke:#047857,color:#fff
    style CX1 fill:#d97706,stroke:#b45309,color:#fff
    style CX2 fill:#d97706,stroke:#b45309,color:#fff

Enable with DISPATCH_MODE=remote in your .env. See the Operator Guide for details.

Risoluto exposes a full JSON API at http://localhost:4000/api/v1/. Here are the key endpoints:

Risoluto stores all config in ~/.risoluto/ (or the directory passed via --data-dir). There is no workflow file to edit — credentials and settings are managed through the WebUI setup wizard or the /api/v1/config/overlay API.

pnpm test                  # Deterministic unit tests (Vitest, 2904 tests)
pnpm run test:watch        # Watch mode for local iteration
pnpm run test:integration  # Opt-in live integration (requires credentials)

The dashboard has a full Playwright E2E suite with 119 smoke tests across 17 spec files and 4 visual regression baselines. Tests run against a Vite dev server with fully mocked API routes — no backend needed.

pnpm exec playwright test --project=smoke   # Smoke tests (119 tests, ~7s)
pnpm exec playwright test --project=visual  # Visual regression (4 baselines)
pnpm exec playwright test --project=visual --update-snapshots  # Regenerate baselines

Full-pipeline integration test that drives the complete Risoluto lifecycle against real Linear + GitHub APIs:

cp scripts/e2e-config.example.yaml scripts/e2e-config.yaml
# Edit scripts/e2e-config.yaml with your credentials and test project
./scripts/run-e2e.sh

Runs 12 phases: preflight, start, setup wizard, issue creation, agent monitoring, PR verification, restart resilience, and cleanup. Produces a structured report in e2e-reports/. See E2E Testing Guide for full setup and config reference.

The default operating mode is intentionally high trust and local-only.

flowchart TB
    S["🎵 Risoluto<br/><sub>Decides WHEN to launch<br/>and WHICH workspace</sub>"]
    C["🤖 Codex<br/><sub>Decides HOW each<br/>turn executes</sub>"]
    P["🌐 Provider / Proxy<br/><sub>Decides HOW the model<br/>call is routed</sub>"]

    S --> C --> P

    style S fill:#2563eb,stroke:#1d4ed8,color:#fff
    style C fill:#d97706,stroke:#b45309,color:#fff
    style P fill:#059669,stroke:#047857,color:#fff

This project draws direct inspiration from OpenAI's Symphony — a framework that turns project work into isolated, autonomous implementation runs. We loved the vision and built our own TypeScript implementation tailored for local, single-host operator use.

MIT

_{Built with ❤️ — Inspired by OpenAI Symphony}