Tangle

Tangle provides isolated sandboxes for running AI agents, recording model calls and tool use, and comparing candidate changes on fresh test cases before human approval. It targets developers and engineering teams in companies of all sizes, from individuals to enterprises. The platform is positioned as infrastructure for running, measuring, and deploying AI agents, with usage-based pricing and a free tier.

Key features

• Isolated sandboxes with shell, files, browser
• Records model calls, tool use, timing, tokens, cost
• Analyzes multiple runs for recurring failures
• Compares candidate changes on held-out cases
• Creates pull requests, issues, skill proposals
• OpenAI-compatible inference API
• Usage-based pricing with free tier