Talk to build software. Describe what you want — an AI team turns it into a structured plan, epics, tasks, and working code.
Website: nocodo.com · Docs & support: nocodo.com
nocodo is a local-first desktop app with a built-in AI team. You describe what you want to build in plain language. Your AI Product Owner asks clarifying questions, hands off to a Project Manager who creates epics and tasks, and specialist agents — backend, frontend, database, UI — execute the work.
No coding knowledge needed. You bring the idea. nocodo brings the team.
Step 1 — Talk to the Product Owner
Start a chat. The PO asks about your users, what data you're tracking, which features matter, and what constraints you're working within. Some questions are free-text; others are click-to-answer widgets for speed. The PO listens and understands — it doesn't create tasks.
Step 2 — PM creates the plan
When the PO has enough clarity, it hands off to the Project Manager. The PM takes the requirements summary and creates a structured epic with tasks, each assigned to the right specialist agent.
Step 3 — Specialists execute
Tasks flow to backend, frontend, database, and UI designer agents. An Engineering Manager reviews technical tasks before they reach specialists. Each task has a clear state: draft → ready → in_progress → done.
Step 4 — Comment and refine
Every epic and task has a comments thread. Specialists, PM, and PO use comments to surface questions and decisions. You stay in the loop without being in the way.
Chat with an AI Product Owner who asks the right questions. Structured choice widgets (radio, checkboxes) keep answers precise and the context clean for downstream agents.
The Project Manager creates epics and tasks from the PO's requirements summary in a single atomic step — no manual decomposition, no copy-pasting notes into a tracker.
Tasks move through a defined lifecycle: draft → needs_technical_shaping → ready → in_progress → done. The Engineering Manager gates technical tasks before they reach specialists. The rules are code, not config.
nocodo maintains two persistent records as your project grows: project notes (business context, decisions, open questions) written by the PO, and stack notes (technical decisions, architecture) maintained by the EM. Future agents read these to stay in context.
Every epic and task has a comments thread. Specialists comment on their assigned work. Users and PM/PO can read and reply. Technical discussion lives next to the work it's about.
Everything runs on your machine. Conversations, artifacts, and agent state are stored in a local SQLite database. No cloud dependency, no lock-in.
nocodo is experimenting with running code generation agents on models under 1B parameters — tiny models that run locally via llama.cpp. The RustEngineer generates Diesel ORM functions by extracting code context from your project, building example-rich prompts, and producing code in a single shot. This means truly private code generation — no API keys, no cloud calls, no data leaving your machine.
| Agent | What they do | Do you talk to them? |
|---|---|---|
| Product Owner | Requirements intake — listens, asks questions, summarises | Yes — this is your main conversation |
| Project Manager | Planning — creates epics and tasks from PO's summary | Yes — in the planning phase |
| Engineering Manager | Technical shaping — reviews tasks before specialists start | Not yet (coming soon) |
| Backend Engineer | Implements backend tasks | Via task comments |
| Frontend Engineer | Implements frontend tasks | Via task comments |
| DB Engineer | Designs and migrates database schemas | Via task comments |
| UI Designer | Produces UI specifications and component designs | Via task comments |
| Rust Engineer | Generates Rust/Diesel ORM code by example | Admin tool — runs locally on tiny models |
nocodo is a Tauri desktop application. The AI backend, agent runtime, and admin UI all run locally.
- Mac, Windows, Linux
- SQLite database on disk — inspect or migrate freely
- No account required to start
- Engineering Manager — fully wired: technical shaping, codebase reading,
needs_technical_shaping → readytransitions - PM edits existing artifacts — follow-up sessions that refine rather than recreate epics and tasks
- Real auth — email/password login; guest sessions that can be claimed
- Expanded code agents — Actix handlers, shared types, SolidJS components generated by small models
- Generated web app — TypeScript + SolidJS + TailwindCSS frontend generated from your project definition
- Backend — Rust + Actix Web
- Desktop — Tauri wrapping the admin UI
- Admin UI — TypeScript + SolidJS + TailwindCSS + DaisyUI
- Shared types — Rust types with TypeScript codegen (no handwritten API contracts)
- Agent runtime — LLM-backed multi-agent system with SQLite-persisted sessions
- Database — SQLite (local default)
- Code agents — llama.cpp with models ≤ 1B parameters for local, private code generation
See DEVELOP.md for local development setup, running the backend and UI, and the deploy workflow.
nocodo is in active development. The two-phase PO → PM chat flow, task state machine, specialist dispatch, comments, and code extraction are all functional. The RustEngineer demonstrates that code generation is possible with models under 1B parameters running locally. Engineering Manager, real auth, and expanded code agent modes are on the roadmap.
Contributions, feedback, and issues welcome at github.com/brainless/nocodo.