# Governor

> A context-driven model router that sits between coding agents and LLM providers to cut token spend by grounding requests in code context and routing to the best model.

Governor is a product from Bito that acts as a single layer between coding agents and the models they call. It routes each request to a model and reasoning level and grounds the request in a live graph of the customer's repositories through Bito's code context engine. Bito announced it on October 27, 2026, and access is currently through a request form or demo.

## What It Is

Governor is a model router for coding agents with a built-in code context engine. Teams keep using Claude Code, Cursor, Codex, Pi, or GitHub Copilot, and point the agent's base URL at Governor. It speaks the Anthropic and OpenAI APIs and runs in or alongside an existing gateway on the team's own provider keys.

## How Routing Works

Before a request reaches a model, Governor classifies it into a tier based on the kind of work and its risk, from carrying out an agreed plan up to planning, code review, and production-touching changes. For each tier it picks a model and reasoning level from the models the team allows, and teams set how aggressive it is on cost. Governor reads each request together with its task, so short messages like "implement this" are treated as the build they are. It makes the decision when a developer starts a new request and holds it while the agent works, and it accounts for what is already cached before switching models. It can also add prompt guidance.

## Code Context Engine

Governor keeps a live graph of repositories down to symbols, calls, and service dependencies. When an agent needs to find something, Governor does the research server-side with a smaller, faster model and returns a short answer citing the file and line. The same context is served to every developer at once rather than only to engineers who install an MCP.

## Cost Visibility and Claims

The product offers an admin view of tokens, spend, and routing decisions, budgets per team or key, and an on/off option to measure savings on rolling traffic. Bito states that in a customer A/B test cost per task fell 48% with task success holding at 100%, and that its own measurements show 60% to 80% of agent cost goes to searching code. These figures are vendor-reported. Bito also states that code is not stored or used for model training and that it can deploy in Bito cloud or on-prem.

## Features
- Context-driven model routing by task type and risk tier
- Selection of model and reasoning level per request
- Prompt guidance added to requests
- Session-aware routing that holds model choice during a task
- Cache-aware model switching
- Code context engine with live graph of repositories, symbols and dependencies
- Server-side code research returning answers with file and line references
- Single base-URL swap compatible with Anthropic and OpenAI APIs
- Runs alongside existing gateways on own provider keys
- Admin view of tokens, spend and routing decisions
- Budgets per team or key
- On/off option to measure savings on rolling basis
- Pinned routes and per-key quality floor
- On-prem or Bito cloud deployment

## Integrations
Claude Code, Cursor, Codex, GitHub Copilot, Pi coding agent, Anthropic, OpenAI, Gemini

## Platforms
WEB, API, VSC_EXTENSION, JETBRAINS_PLUGIN, CLI

## Pricing
Paid

## Links
- Website: https://bito.ai/product/governor
- Documentation: https://docs.bito.ai/
- EveryDev.ai: https://www.everydev.ai/tools/bito-governor