Blog
Tangle builds infrastructure for AI agents. These guides cover sandboxes, the agent runtime, model routing, browser agents, evaluation, and traces. Each series reads in order. New to Tangle? Start with Introduction to Tangle.
Introduction to Tangle
Tangle gives each agent run its own sandbox, runs the agent loop, routes its model calls, and records each step as a trace.
Agent Runtime Infrastructure
How an agent runtime keeps one execution path and rejects a worker budget too small to finish.
The Instrument Problem
How to keep coding-agent benchmarks and behavioral signals aligned with the decision they support.
When Structure Pays
A measured comparison of one agent and repeated candidates, including quality, token cost, elapsed time, and uncertainty.
Building an AI Tax Agent
A proposed tax preparation workflow where every number links to its source document and a person approves the return before it is filed.
Code Auditor
How an agent turns a scanner alert into a finding another reviewer can reproduce, with reachability and impact evidence.
Browser Agent
Browser tests written as English goals that keep screenshots and page state, check the final result, and stop before irreversible actions.
Blueprint Agent
Blueprint Agent is a browser coding workspace; this post checks hackathon builds by running the code.
Agent Intent Infrastructure
How an agent discovers a product and how to pick a browser agent.
The Self-Improving Stack
When to change an agent's prompts, the code around the model, or the model weights, and the evidence each change needs before it ships.
Standalone
The Tangle Network protocol is a separate project: Blueprint SDK Guide: Run Rust Services on the Tangle Network, Tangle Network Basics, Tangle Network Protocol, x402 Production Runway.