Plan → run → audit → review

aidd

A local control panel for planning AI coding work and checking whether it really passed.

Point it at a codebase, give it a spec, and choose the agent or model you already use. aidd builds the backlog, runs bounded agent sessions, keeps features open until the gates pass, and leaves behind the run history you need later. In current market language it is a local, agent-agnostic agentic development environment: delegated work and durable results are the primary units, not files in an editor. Your projects, your keys, your machine.

Get latest source View source

Every capability included
No account
No telemetry sent. No phone home.

Backends: Claude, Codex, local
Control: Time, iterations, budget
Verification: Tests, lint, browser gates
Cost: Free, every capability

aidd was built out through seven creation lanes across three languages, with real gates, published failures as well as passes, and run ledgers you can inspect.

See receipts Read changelog

The field

The missing piece is usually verification.

In 2026 you can rent a cloud agent by the task, put a board over your terminal agents, or generate a clean spec. Those are useful pieces. aidd is for the awkward part after that: turning the spec into work, running the agent under limits, and making the result pass the project gate.

Use aidd when

You need to trust an outcome, not merely watch a terminal.

Your work spans real repositories, repeatable workflows, quality gates, and human approval, on one project or across a fleet.

Skip aidd when

You want autocomplete, a replacement IDE, or one-click unattended autonomy.

aidd is not the coding agent. If you do not need an evidence trail, it is more process than you need.

Cloud agents

Hosted execution

Hosted agents make it easy to hand off work without keeping a local machine running. Pricing, sandbox access, and the evidence returned with a run vary by provider.

aidd: uses the agent subscription you already have, on your machine, under iteration and wall-clock caps with the run cost reported as it accrues.

Agent boards

Session visibility

Kanban boards, worktree apps, and terminal managers provide useful views of active sessions and diffs. Planning depth, automated gates, and platform support depend on the product and configuration.

aidd: keeps a feature open until tests, typecheck, lint, and browser verification pass.

Spec kits

Requirements first

Spec-driven kits help turn an idea into structured requirements. Execution and verification are usually handled by the agent workflow that consumes those artifacts.

aidd: keeps the spec connected to dependency-ordered work, acceptance criteria, gate checks, and a decision queue for the calls a tool should not fake.

What you need	Cloud agents	Agent boards	Spec kits	aidd
Where the work runs	Provider-managed environment	Your machine	Your repo	Your machine, your keys
Spec becomes a backlog	Task or issue based	Often user-managed	Requirements and planning artifacts	Dependency-ordered, with acceptance criteria
“Done” means	Provider workflow completed	Project-defined review	Requirements are ready for execution	Tests, typecheck, lint, and browser gates passed
Cost model	Per-task metering	Free, futures vary	Free CLI	aidd is free; your existing agent plan, metered per run
Run receipts	Provider-dependent	Session history	Planning artifacts	Commits, iterations, tool activity, cost
Windows	Browser-accessible	Product-dependent	Varies	Smoke-tested, first-class

High-level category snapshot, July 2026, rather than a feature-by-feature comparison. Examples include Devin, OpenAI Codex cloud, Google Jules, Copilot coding agent, Factory, Vibe Kanban, Conductor, Sculptor, Claude Squad, GitHub Spec Kit, AWS Kiro, and BMAD. Capabilities and pricing change; aidd is focused on the local planning, run, and verification loop.

The loop

From folder to verified feature.

aidd turns intent into backlog items, selects runnable work, launches bounded agent sessions, checks the result, and parks anything that needs a human call.

Ingest the codebase

Add application roots in Settings. Any folder with an .aidd/ contract becomes a project, including foreign stacks.

Plan the backlog

Onboarding turns a plain-language spec into dependency-ordered feature work with verifiable acceptance criteria.

Launch gated runs

Use Claude Code, Cline, Codex, Grok, OpenCode, or KiloCode, or OpenAI-compatible endpoints including OpenAI, LM Studio, and Ollama.

Prove the work

Features complete only after the project gate passes: tests, typecheck, lint, and browser-level verification when the project has a UI to drive.

Review receipts

Every run keeps commits, iterations, tool activity, cost, and an honest summary. Unclear work lands in the decision queue.

Receipts

The proofs used real apps.

aidd built and repaired real applications through seven different creation lanes. The runs were there to find rough edges, file the failures, fix them, and leave a public trail.

All seven are published in full - app snapshots, run ledgers, iteration evidence, and scrubbed transcripts - in the build-proofs repository, including the lane that failed. This is first-party evidence, not a benchmark or an independent audit.

Lane	App	Stack	Result
Fresh scaffold from a spec	Habit tracker	Bun + TS + React	25/25 features passing, gate clean, roughly 2.5 h
Third-party template init	Kanban board	Vite + React	23/23 features, 10/10 headless acceptance scenarios in the gate
Full-stack template init	SMB infrastructure dashboard	Bun + Elysia + React + Drizzle	26 features passing plus one retained remediation record, 30-step gate green
Existing-codebase ingest	Flask tutorial app	Python + Flask	Profiled the foreign stack, produced a grounded security audit, then fixed findings with pytest as the gate. Zero writes outside `.aidd/`.
GitHub-template clone	Pantry tracker	PowerShell + Pode + htmx	26/26 features, Pester 88/88, analyzer clean
Local model only	Markdown TOC tool	Bun + TS, LM Studio, no cloud calls	0/14 features. Five coding runs finished nothing, and the completion contract refused to record success. Published exactly as it ran.
Agent-driven build	Margin calculator	Bun + Elysia + React + Drizzle	29/29 features across 48 supervised iterations, gate green at the end

Trust model

Built for people who let agents touch real repos.

The default posture is deliberately plain: local data, no account, explicit model backends, bounded runs, and documented write surfaces.

Local

Your machine stays in charge

The runtime, panel, ledger, and project metadata live locally. Network traffic goes only to the model backends you configure.

Quiet

No account, no telemetry

There is no aidd account and no hosted control plane. It sends no telemetry and does not phone home; network traffic comes only from providers or integrations you configure, or downloads you initiate. The write allowlist is documented, including what lives under .aidd/ and when source can change.

Bounded

Hard limits on runs

Iteration caps, wall-clock caps, dirty-tree guards, and parked decisions stop a run outright; cost and token budgets are warn-only, reported against the run rather than cutting it off mid-edit.

Open

Security docs in public

Review SECURITY, PRIVACY, and SUPPORT.

Included under FSL

Every capability is included.

Run aidd supervised or unattended, on one project or across a fleet. Every capability ships together under FSL. You provide access to the AI providers you choose.

aidd

One product

Everything aidd does ships together, for any number of projects and machines.

Runs and gates
All backends
Audits
Recipes and skills
Director cycles
Scheduled fleet automation
Continuous follow-ups
Telegram bridge
MCP server

Source-available under FSL-1.1-ALv2, converting to Apache 2.0 two years after each release.

Install from source

Project state: Your local .aidd/ metadata
Control plane: Your machine
aidd account: None
Telemetry sent: None
Model access: Your configured provider or agent CLI

Get it

Four commands. Nothing to sign up for.

Windows 11 + PowerShell 7 is the current smoke-tested path. Install Git and Bun 1.4.0 or newer, then run aidd directly from its source checkout.

Source install

git clone https://github.com/NomadicDaddy/aidd.git C:\Tools\aidd
Set-Location C:\Tools\aidd
bun install
bun run start:web

# Open http://127.0.0.1:3210

aidd ships under FSL-1.1-ALv2: available for any use except making it available to others as a competing commercial product or service, and each version automatically becomes Apache 2.0 two years after release. The .aidd/ contract format is an open spec anyone may implement. Spernakit, the companion full-stack template, is MIT.