Autohand AI
Autohand AI builds infrastructure for autonomous software: AI coding agents and governed software-factory systems that write, test, review, deploy and continuously improve software. Its stated goal is Level 5 autonomous programming, where software identifies needs, implements solutions and evolves with minimal human intervention.
At a Glance
- Regulated enterprises
- Software engineering teams
- Financial services
- Healthcare
- +5 more
AI Tools by Autohand AI
(2)
Zit
Git Extension for Parallel Agents
Autohand Code
Open Source Terminal AI Agent
Discussions
No discussions yet
Be the first to start a discussion about Autohand AI
Latest News
Published “When to Build Custom MCP Servers: A CTO Decision Framework,” covering control, cost, sensitive integrations and deployment rules for community versus custom MCP servers.
Published product and engineering updates including “Autohand Model Migration Checklist” and “MCP Integration Checklist.”
Introduced Zit, a Git extension for coding agents working in parallel, using copy-on-write workspaces, claims and transactional acceptance checks.
Introduced Notebooks in Autohand Code Web for exploring connected data through a single data-engineering and data-science entry point.
Products & Services
Open, composable AI coding agent for terminal, web, desktop, VS Code, Zed, JetBrains, Chrome, Slack and SDK workflows. It plans, edits, tests and reviews code, understands whole codebases, orchestrates agent teams and can execute autonomous tasks.
Self-evolving AI software-engineering system for measurable engineering problems. It generates competing candidates, evaluates them against a scorer, preserves evidence and promotes improvements through an auditable search loop.
SDK and API for building review bots, CI agents and internal developer tools on the Autohand agent loop; supports streaming events, self-hosting or cloud use and 9+ languages including TypeScript, Python, Go, Java, Swift, Rust, Ruby, C# and C++.
Open-source Git extension, sponsored and maintained by Autohand AI, for multiple developers and coding agents changing one repository concurrently. It uses copy-on-write workspaces, claims, stale-change checks, checks declared in zit.toml and plain Git objects.
Market Position
Autohand positions itself beyond autocomplete and conventional human-centric coding assistants as an open, composable and privacy-first autonomous software factory. Its founder explicitly contrasts the machine-orchestration focus with Cursor, Claude Code and OpenAI Codex, while the company differentiates through whole-codebase understanding, multi-agent execution, governed enterprise delivery and Evolve's measurable self-improvement loops.
Leadership
Founders
Igor Costa
Founder and CEO; previously held senior engineering roles at AWS and GitHub/Microsoft, where he helped deploy GitHub Copilot to millions of developers.
Executive Team
Igor Costa
Founder and Chief Executive Officer
Former AWS and GitHub/Microsoft engineer; helped deploy GitHub Copilot to millions of developers.
Board of Directors
Founding Story
Igor Costa says he left GitHub in May 2025 after deciding to build the future of software development rather than watch someone else build it. Having seen how developers worked at AWS and GitHub/Microsoft, he wanted to build the infrastructure or factory floor for software industrialization—not merely an incremental autocomplete tool—with privacy, configurability, modularity and machine-scale orchestration.
Business Model
Revenue Model
Freemium and paid monthly subscriptions for hosted Autohand Code capacity, with per-seat team/enterprise plans; BYOK and local inference are supported, while enterprise offerings and Evolve are sold through contact-sales/demo motions.
Pricing Tiers
15 credits per rolling 5-hour window; up to 200 per month; includes Fantail, BYOK, SDK/API, open-source CLI, code generation/web search and editor/mobile access.
125 credits per rolling 5-hour window; up to 500 daily and 2,500 weekly.
250 credits per rolling 5-hour window; up to 1,000 daily and 5,000 weekly.
1,000 credits per rolling 5-hour window; up to 4,000 daily and 20,000 weekly.
3,000 credits per rolling 5-hour window; up to 12,000 daily and 60,000 weekly.
3,000 credits per member per rolling 5-hour window; up to 12,000 daily and 60,000 weekly per member.
1,200 credits per member per rolling 5-hour window and 20,000 weekly per member; agreements can add custom capacity.
Target Markets
- Regulated enterprises
- Software engineering teams
- Financial services
- Healthcare
- Telecommunications
- Defence and national security
- Code modernization and legacy migration
- Technical-debt reduction and autonomous SRE
- Regulated-enterprise software delivery with policy controls and audit trails
- CI/CD refactoring, code review, testing and security workflows
- Long-horizon engineering problems with measurable evaluators, benchmarks or simulations
- Parallel coding-agent collaboration on one repository
- Microsoft
- Xero