SEMANTIC DEV INTELLIGENCE

Ship enterprise software
the work of ten engineers,
driven by one.

MetaCoder reads your whole codebase as a Semantic Knowledge Graph...

Windows · macOS · Linux · Free to start · Bring your own API key

metacoder — console
# MetaCoder Desktop · Harness Engineering
 
/modernize --output ./web
Phase 0 · building knowledge graph …
graph 412 entities · 86 routes · 19 services
Phase 4 · multi-agent implementation
gate-implementation typecheck · unit-test
Phase 5 · systest (Chrome MCP)
28/28 E2E evidence captured
MODERNIZATION_REPORT.md
10
Pipeline phases
requirements → confirm
4
Built-in roles
PM / Dev / Reviewer / Tester
14
Quality gate types
machine-verified
3
Context layers
session / phase / on-demand
BUILT FOR SCALE

Engineered for 10,000+ file codebases

10,000+
Whole-system
Zero blind spots
YOU THINK, METACODER BUILDS

Stay on the ideas. Delegate the build.

You focus on

  • System planning
  • Requirements & the product idea

MetaCoder handles

  • Architecture & design
  • Coding & implementation
  • Testing & QA
  • Database setup & schema
  • Cloud infra design & config
THE CORE PHILOSOPHY

Harness Engineering

AI writes code fast — but "fast" without order produces drift...

KEY INNOVATION · SPEC-KIT

spec-kit — design first, then build exactly to the design

Without spec-kit
With spec-kit

A state-machine pipeline

Roles that switch automatically

Honest-fail gates

Anti rubber-stamp review

Cross-cutting capture

Full audit trail

The 10-phase pipeline

Each step runs a gate before it can advance.

01
Requirements
GATE
02
Design
GATE
03
Test Spec
GATE
04
Implementation
GATE
05
Code Review
GATE
06
Integration Test
GATE
07
Performance
GATE
08
Security
GATE
09
Deploy Plan
GATE
10
Confirm
✓ DONE
↩ rollback Same pipeline for humans and AI.

/harness — the command reference

Setup & inspect
/harness init [--template TYPE]
/harness spec [--inputs PATHS] [--advance] <brief>
/harness validate
/harness lint
/harness status
/harness phase init [--feature NAME]
Pipeline control
/harness phase status
/harness phase advance [--reason]
/harness phase rollback [--reason]
/harness gate <gateId>
/harness assume <role>
/harness role
Implement & advanced
/harness implement
/harness context <show|refresh|query>
/harness approve <gateId>
/harness drill [--from --to]
/harness ai-rate [--since N]
/harness ci-init [--target]
AUTOMATION SKILLS

Four super-tools, one engine

Every skill builds on the same Semantic Knowledge Graph...

LEGACY → MODERN WEB

/modernize

REQUIREMENTS → APP

/newproject

AUTO QA · CHROME MCP

/systest

INFRA · SAFETY RAILS

/env

Command deep-dive

$ /modernize --workspace ./legacy --database <conn> [--output ./web] [--design-style] [--backend-lang] [--frontend-lang] [-v] [--env]
Phase flow
Highlights
$ /newproject --workspace ./app --design-docs ./specs [--database <conn>] [--reference] [--design-style] [-v] [--env]
Phase flow
Highlights
$ /systest run --workspace ./app [--backend-url <url>] [--frontend-url <url>] [--database <conn>] [--env]
Phase flow
Highlights
$ /env <list|use|add|edit|remove|clone|set|import|export|doctor|diff>  full reference →
Subcommands
Highlights
HOW IT WORKS

Your code becomes intelligence the AI can reason over

INPUT
Source code
GRAPH
Knowledge graph
OUTPUT
Grounded AI work
DATABASE & CLOUD, HANDLED

TiDB Serverless, built into the Desktop app

One-click TiDB Serverless

Cloud infra by MetaCoder

Zero hand-wiring

TWO WAYS TO RUN

Console & Desktop — same engine

BRING YOUR OWN MODEL

Works with the APIs you already use

Claude
DeepSeek
Zhipu GLM-4
OpenAI-compatible

FAQ

Frequently asked questions

What is MetaCoder?
What is Harness Engineering?
Which AI models and APIs does MetaCoder support?
Can MetaCoder handle very large codebases?
How do I get MetaCoder, and does it set up a database?
GET STARTED

Download MetaCoder

Latest: v3.6.22 · 2026-06