Aigon GitHub

Spec-driven development · Multi-agent orchestration

Spec-driven AI development.
Orchestrate any agent — or all of them at once.

Aigon captures the full product lifecycle — research, feature delivery, and user feedback — in specs and logs committed directly to your repository. Run one agent with tight control, or orchestrate competing implementations in parallel — then evaluate and merge the best outcome.

Start in 5 minutes How it works
  • Four clear modes: Drive, Fleet, Autopilot, and Swarm
  • Specs, logs, and research findings in Git — searchable, portable, permanent
  • Works with Claude, Gemini, Cursor, and Codex — swap freely or run them head-to-head
  • No required SaaS account, plain files in Git, no proprietary formats

The problem

AI agents multiply output fast, but coordination breaks down even faster.

Terminal tabs everywhere

Parallel agents generate momentum, but tracking what each one did and why becomes manual overhead.

Decisions get lost

Without spec-linked workflows, merge decisions and implementation tradeoffs disappear into chat history.

No clean comparison step

Teams rarely compare multiple implementations side-by-side before merging the best approach.

Feedback loops stay disconnected

User feedback, research, and implementation often live in separate tools with no shared lifecycle.

Value proposition

Aigon turns multi-agent output into an auditable delivery system.

Traceability

Specs and decisions live in your repo

Research, feature specs, implementation logs, and evaluations stay in Markdown files your team can review and version.

docs/specs/features/03-in-progress/
docs/specs/features/logs/

Vendor independent

Use the agents you already prefer

Run the same workflow with Claude, Gemini, Cursor, and Codex without rewriting your process around one tool.

aigon install-agent cc gg cx cu

Shared lifecycle

Research, delivery, and feedback connect end to end

Aigon links discovery, implementation, review, and follow-up so each cycle improves the next one.

research -> feature -> eval -> close

Operational clarity

Mode-based commands reduce team confusion

Pick the right execution mode for each task and make expected behavior explicit before coding starts.

aigon feedback-triage 14

How it works

A continuous loop: research, build, review, learn.

Aigon structures the full product lifecycle as specs and logs committed to your repository. Research feeds features, features feed implementation, implementation feeds evaluation, and user feedback feeds the next cycle — one auditable system from discovery to done.

State-as-location

Task status is represented by folder location, making progress auditable in Git and obvious in code review.

01-inbox 02-backlog 03-in-progress 04-done

Choose your mode

Hands-on or hands-off, one agent or many.

Hands-on + one agent

Drive mode

Use when you want tight control over implementation details and review checkpoints.

aigon feature-setup 07

Outcome: one guided implementation branch with a full spec and log trail.

Hands-on + multi-agent

Fleet mode

Orchestrate competing implementations you can evaluate and adopt the best from.

aigon feature-setup 07 cc gg cx

Outcome: parallel worktrees and comparable outputs for structured selection.

Hands-off + one agent

Autopilot mode

Use when the scope is clear and you want autonomous retries against validation checks.

aigon feature-autopilot 07

Outcome: automated implement-validate loop that stops when checks pass or budget ends.

Hands-off + multi-agent

Swarm mode

Fully orchestrated, fully autonomous — parallel agent runs converge into comparable outputs ready for evaluation.

aigon feature-setup 07 cc gg cx
# in each worktree:
aigon feature-autopilot 07 --auto-submit

Outcome: concurrent autonomous runs across agents with logs ready for comparison.

The steps

Four explicit steps from idea to merged code.

01

Define

aigon feature-create "jwt-auth"
aigon feature-prioritise jwt-auth

02

Set up mode

# Drive
aigon feature-setup 07
# Fleet
aigon feature-setup 07 cc gg cx
# Swarm
aigon feature-setup 07 cc gg cx
# then in each worktree:
aigon feature-autopilot 07 --auto-submit

03

Do

aigon worktree-open 07 --all
# then in each agent:
/aigon:feature-do 07

04

Evaluate, merge, and adopt

# in your agent:
/aigon:feature-eval 07
aigon feature-close 07 cx --adopt
aigon feature-cleanup 07

If any agents haven't submitted yet, eval warns you before proceeding — so you never accidentally compare incomplete work.

See it in action

One canonical command path for each mode.

Inside an agent

Slash commands in a live agent session

From the CLI

Shell commands for autonomous execution

Your workflow, visualised

A Kanban board for spec-driven development.

The Aigon Dashboard is the visual way into your spec-driven workflow. Same pipeline, same agents — but managed through a browser UI instead of CLI commands. Drag features across columns, launch agent sessions with one click, and watch your pipeline move in real time. No terminal required.

Aigon Dashboard Kanban board showing feature cards in inbox, backlog, in-progress, evaluation, and done columns

Visual Workflow

Drag specs from inbox to done

Move features through your development pipeline with a familiar Kanban interface. Every column maps to an Aigon workflow state — no commands to memorise.

Monitor

See every agent, every repo, at a glance

The monitor tab shows running sessions, attention items, and recent events across all your repositories — everything you need to know at a glance.

Measure

Throughput, cycle time, agent performance

The statistics tab turns your spec history into metrics. Know which agents ship fastest, how your cycle time trends, and whether your pace is accelerating.

Documentation

Install in minutes, then run your first end-to-end loop.

git clone https://github.com/jayvee/aigon.git
cd aigon
npm install
npm link
cd /path/to/your/project
aigon init
aigon install-agent cc gg cx cu
aigon feature-now "dark-mode"

Workflow Guide

Research, specs, implementation, evaluation, and completion flow.

CLI README

Install instructions, command reference, and real examples.

Tool Comparisons

Where Aigon is stronger, where it is weaker, and how to combine tools.

Tech & philosophy

Open source, git-native, and intentionally simple.

Aigon is built for teams who want disciplined AI-assisted engineering, not opaque automation.

  • Open SourceMIT licensed, no paid-tier lockouts.
  • Repo-Native ContextSpecs, logs, and evaluations stay as plain Markdown in your repository.
  • Agent-AgnosticWorks with whichever coding agents your team chooses.
  • Adapts to Your StackWorkflow templates and defaults adjust for web apps, APIs, iOS, Android, and libraries.

Community

Help shape the next generation of collaborative AI development.

Contribute specs, improve workflows, and share real-world patterns for running multi-agent engineering teams effectively.

Star on GitHub Install guide