Now in private beta ยท 247 specs passing

Ship software
autonomously.

A self-improving multi-agent factory. Submit a spec. Dark Factory researches, plans, codes, tests, and validates โ€” iteratively, until it converges.

Get started free Sign in
SOC2-ready architecture
Tenant-scoped encryption at rest
Blue-green deploys
Real-time cost attribution
0
LLM providers supported
0
Pipeline templates
0
Agent handlers
0
Specs passing
How it works

From spec to working code
in four steps.

01

Submit a spec

Describe what you want in plain English. Reference colors, edge cases, test scenarios โ€” or just a one-line feature idea.

02

Agents plan

Research, contract, and harness-gen agents translate your spec into a DAG of work: requirements, test scenarios, file layout.

03

Build & test

Developer, tester, infra, and build agents execute in parallel streams. Every artifact is validated against the harness.

04

Converge or retry

Critic reviews the output. If satisfaction < threshold, the factory learns from the trace and iterates with new context.

Architecture

A DAG of autonomous agents.

Templates are .dot files describing work as a graph. Handlers self-register. The engine walks the DAG, passes state, and re-runs failed nodes with retry logic.

seed research contract harness_gen codergen build_test critic_review harness exit
31 handlers 7 templates Plugin registry Automatic retry + convergence
Capabilities

Everything you need to ship.

From spec to production โ€” multi-tenant, observable, audited.

Multi-agent pipelines

DOT-based DAG templates. 31 self-registering handlers. Parallel streams for engine and UI. Plugin architecture.

Self-improving

Skill bank, instinct engine, knowledge graph. Agents learn from every trace. Pattern recognition across runs.

5 LLM providers

DeepSeek, Anthropic, OpenAI, MiniMax, Dashscope. Tenant-scoped encrypted credentials. Bring your own keys.

Cost attribution

Per-call token ledger with raw + billed cost. Pricing plans. Markup overrides. CSV export for finance.

Multi-tenant SaaS

Devise auth with invitations. Role-based access (owner/admin/member). Tenant-scoped queries throughout.

7 pipeline templates

iOS, Android, Rails API, fullstack web, web frontend, Ruby library, multi-platform. All customisable as .dot files.

Context management

Smart compaction, tool filtering, per-turn budget. Context overflow recovery. Handles 100k+ token prompts.

Deterministic infra

Auto-generates Package.swift, Gradle, Bundler, setup.py from workspace contents. No LLM round-trips.

Real build validation

Every iteration actually compiles and tests. Swift, Kotlin, Ruby, Python โ€” with failure loop and retry.

Critic review

Dual-LLM validation: one builds, another reviews. Decision tree chooses when to iterate vs ship.

API + browser UI

Bearer-token API v1. Dashboard with live streaming runs. Turbo Streams. Export everything.

Enterprise deploys

GitHub Actions CI/CD. Blue-green Kamal deploys with automatic rollback. pgvector. Solid Queue. kamal-proxy.

Pricing

Pass-through pricing.

You pay the real LLM provider cost. We add a markup you can see on every ledger entry.

Starter
0%
markup

Pass-through pricing. No markup.

  • Bring your own LLM keys
  • All 7 pipeline templates
  • Factory run dashboard
  • CSV export
  • Community support
Start with Starter
Most popular
Pro
20%
markup

Team features with sensible markup.

  • Everything in Starter
  • Tenant invitations
  • Role-based access
  • Priority CI runs
  • Email support
Start with Pro
Enterprise
50%
markup

Dedicated support, custom SLAs.

  • Everything in Pro
  • Custom markup override
  • SSO (on request)
  • Custom handlers
  • Dedicated Slack channel
Start with Enterprise

Ready to submit your first spec?

Create an account, add an LLM credential, and watch Dark Factory ship working code in under five minutes.

Get started free Sign in