Route between coding-agent models with the open-source Weave Router 2.0
The Weave team presented Weave Router 2.0 in a Show HN post. It is an open-source routing model that sits in front of coding agents and switches between LLMs per turn, and the authors say it matches the pass rate of a single frontier model at lower cost and with faster completion.
The problem
Coding agent sessions can involve around 100 LLM calls, and each call is a chance to pick a different model. Using one top model for everything is expensive, and switching models carelessly wastes money because each switch means filling a different prompt cache. The team wanted a router that could pick the right model per turn without exploring every possible path.
How they did it
- Plug the Weave Router into a coding agent harness such as Claude Code, Codex, OpenCode or Pi.
- Choose between self-hosting the open-source router from GitHub and the hosted version at weaveos.com/router.
- For self-hosting, supply provider access; OpenRouter is suggested as an easy start but is not required.
- Let the router send hard tasks such as tricky debugging to a strong model and simple frontend changes to a cheaper one.
- Optionally set a predefined budget, which the author confirmed is supported.
Results
- As reported by the Weave author on Hacker News: on Terminal Bench 4.0 and SWE Atlas, the router had equivalent pass rates to GPT-6 Astra.
- As reported by the author: on Terminal Bench the router cost 52% as much as Astra and completed tasks 2.2x faster.
- As reported by the author: on SWE Atlas the router cost 54% as much as Astra and ran 2.5x faster.
- The author credits a hidden Markov model plus classifier for bucket selection as the biggest gain, and cache-eviction impact estimation for most of the cost savings.
As reported by the source (Hacker News Show HN post by the Weave Router author); AgentGid did not measure these figures.
This suits developers or small teams already running a terminal coding agent with provider keys, and comfortable self-hosting a router or trying the hosted one. Watch the figures: the pass-rate and cost numbers come from the author, and your savings depend on provider pricing and your own task mix. Pi (Free, OSS) is a no-cost harness to pair it with, since you only pay the model provider.
The agents used here
Anthropic's agentic coding tool that reads a codebase, edits files and runs commands from the terminal, with VS Code and JetBrains integrations plus …
- Price
- $20/mo
- Gid Score
- 80 #2 in Terminal
OpenAI's coding agent, offered as an open-source terminal CLI, an IDE extension, a desktop app and a cloud agent that works on repositories. Users si…
- Price
- Free + $8/mo
- Gid Score
- 83 #1 in Terminal
Open-source coding agent from Anomaly (the team behind SST) that runs as a terminal UI, desktop app or IDE extension and connects to 75+ model provid…
- Price
- Free + $10/mo
- Gid Score
- 74 #4 in Terminal
Minimal, extensible open-source coding agent harness with interactive TUI, print/JSON, RPC and SDK modes, created by Mario Zechner and now maintained…
- Price
- Free (OSS)
- Gid Score
- 75 #3 in Terminal
Compare: Claude Code vs OpenAI Codex · Claude Code vs OpenCode · Claude Code vs Pi · OpenAI Codex vs OpenCode
Similar use cases
Run Claude Code and Codex agents in parallel on Mac with Offrun
Offrun is a free Mac app that runs coding agent CLIs such as Claude Code and Codex side by side, each in its own git worktree, using the user's own logins. The page is the product…
As reported by the Offrun site: each agent in a project works in its own git worktree, so two agents on the same repo do not touch the same files.
Run the pi harness in isolated sandboxes with the self-hosted pi pod
An independent developer, Evan, introduces pi pod, an ongoing open project built on top of the pi coding harness. It aims to run a user's existing pi setup in an isolated sandbox …
No measured results are reported; the article is a project introduction.
Review every pull request with Claude Code in GitHub Actions
A GitHub Actions workflow runs Claude Code whenever a pull request is opened or updated. Claude leaves inline comments on the problems it finds, and anyone can also call it from a…
Claude leaves an inline comment on each issue it finds, or a single summary comment when it finds nothing.
Guides
Best AI Agents in 2026: Live Rankings by Category
The leading AI agents for coding, design, research, data, marketing, sales, support, HR, finance, legal, DevOps and security, ranked by public usage data that refreshes daily, with prices and free options.
What Is an AI Agent? A Practical Explanation With Real Examples
What makes software an AI agent, how agents differ from chatbots and plain language models, the main types in use today and how to judge one.
AI Agent Statistics 2026: Live Usage, Growth and Activity Data
Current statistics on AI agents from public data: package downloads, editor installs, GitHub stars, pull requests opened by coding agents, release pace and benchmark results. Updated daily.
Best AI Coding Agents in 2026: Ranked by Usage, Benchmarks and Price
The AI coding agents developers use most, with live download and install numbers, benchmark results, entry prices and a way to choose between terminal, IDE and cloud agents.