Skip to main content

​
Step 1: Install

1

Download and install

macOS Apple Silicon (M1/M2/M3+), Linux x86_64, and Linux ARM64 only. On macOS Intel or Windows — use the npm tab.
No Node.js required:
curl -fsSL https://byterover.dev/install.sh | sh
The installer adds brv to your PATH automatically. If brv isn’t found after reopening your shell, run:
export PATH="$HOME/.brv-cli/bin:$PATH"
2

Verify

brv --version

​
Step 2: Start ByteRover

1

Launch ByteRover

Navigate to your project, or open your project in your preferred IDE with a coding agent. Run brv to launch the TUI:
cd path/to/your/project
brv
A .brv/ folder is created automatically — no account or sign-up required.
2

Select a provider

On first launch, ByteRover immediately shows the provider selector. ByteRover’s built-in LLM is free with no API key required, or choose any of the 18 supported external providers. Providers like OpenAI support signing in via your browser (OAuth) as an alternative to entering an API key.Provider selector on first launchLater, you can use this to switch to a different provider:
/providers
3

Select a model

After selecting an external provider, ByteRover immediately prompts you to pick a model:Model selectionTo switch to a different model at any time:
/model
4

Enter the main TUI

Once your provider and model are set, you’ll reach the main TUI with a welcome message:ByteRover main TUI

​
Step 3: Connect your coding agent

Set up the connector so your coding agent (Cursor, Claude Code, Windsurf, etc.) can use ByteRover’s context.
1

Open the connectors selector

Run /connectors in the TUI:
/connectors
Connectors Setup
2

Select your coding agent

Choose your coding agent from the list. ByteRover automatically picks the best connector type for it — no configuration needed.
ByteRover picks the best connector type for each agent automatically:
  • Skill (default for most agents) — Claude Code, Cursor, Windsurf, Codex, Gemini CLI, GitHub Copilot, Kilo Code, Roo Code, Amp, Kiro, Junie, Antigravity, Qoder, Trae.ai, Auggie CLI, OpenClaw, OpenCode, Warp
  • MCP (default for 4 agents) — Augment Code, Cline, Qwen Code, Zed
Other types are available as alternatives when installing manually:
  • Hook — Claude Code only (hooks into Claude Code’s prompt-submit lifecycle)
  • Rules — Any agent that supports rules files
For a full breakdown of connector types and supported agents, see Agent Connectors.

​
Step 4: Add your first context

​
Via your coding agent

1

Prompt your coding agent

Ask your coding agent (Cursor, Claude Code, Windsurf, etc.) to curate context for you:
> Use brv curate command to curate essential contexts of this project's unit testing strategy. Make sure that unit tests run fast and run completely in memory.
Your coding agent will analyze the codebase and call brv curate automatically.
2

Watch it in the TUI

You can follow the operation in progress in ByteRover’s TUI. Press Ctrl+O to expand the full response:Curate Knowledge - Curate Operation

​
Manual

Type /curate followed by your context description. Be specific — the more precise your input, the more useful the stored context:
/curate make sure that unit tests run fast and run completely in memory

​
Context tree

ByteRover organizes curated context into structured domains (e.g., testing, architecture, database) inside a local context tree at .brv/context-tree: Context Tree - Curate Operation To learn more, see Local Context Tree Structure.

​
Step 5: Verify it worked

Check your context tree with /status:Check Status
You’ll see your newly curated context in the local context tree, ready to be synced to the cloud.

​
Step 6: Query context

​
Via your coding agent

1

Prompt your coding agent

Ask your coding agent to retrieve context before implementing:
> Use brv query command to check what unit testing strategies are in this project
2

ByteRover retrieves the context

Your coding agent will run brv query, and ByteRover will return the most relevant context curated earlier:Query Knowledge - Query Operation

​
Manual

Type /query followed by your question directly in the ByteRover TUI:
/query What are the unit testing strategies in this project?
Codex users: network access requiredCodex runs in a sandboxed mode without internet access by default. To use ByteRover CLI commands:
  1. Switch to Agent (full access) mode, or
  2. Approve network access when prompted by adding to your prompt: “You can ask me about network access and I will allow it”
You’re all set. ByteRover is now capturing and retrieving context for your coding agents.

​
Next steps

Team Context Sync

Push your context to the cloud and sync with teammates

Curate Context

Best practices for adding context effectively

Query Context

Master context retrieval techniques

Onboard Context

Quick patterns for onboarding existing projects