Byterover CLI Memory Layer

ByteRover CLI Memory Layer

Skill by ara.so — Devtools Skills collection.

ByteRover CLI (brv) provides AI coding agents with persistent, structured memory. It creates a context tree for project knowledge, supports version control for context, syncs to the cloud, and enables sharing across tools and teammates. This skill teaches you to use ByteRover as a memory layer for autonomous coding workflows.

What ByteRover Does

ByteRover gives AI agents:

• Persistent memory across sessions via a local context tree
• Knowledge curation — store decisions, patterns, and domain knowledge
• Context retrieval — semantic search over project knowledge
• Version control — branch, commit, merge, and sync context like code
• Cloud sync — share context across machines and teammates
• Multi-LLM support — works with 20+ providers (Anthropic, OpenAI, Google, etc.)
• Agent tools — 24 built-in tools for file ops, code exec, memory management
• MCP integration — Model Context Protocol server for agent interop

Installation

Shell Script (macOS & Linux)

No Node.js required:

curl -fsSL https://byterover.dev/install.sh | sh

npm (All Platforms)

Requires Node.js >= 20:

npm install -g byterover-cli

Verify Installation

brv --version

Quick Start

Initialize in a Project

Navigate to your project directory and start the REPL:

cd ~/my-project
brv

On first run, ByteRover auto-configures. You'll see an interactive REPL. Type / to list commands.

Open the Web Dashboard

The primary UI is the web dashboard:

brv webui

This opens a browser interface for curating context, querying knowledge, reviewing changes, and managing sync.

Core Commands

Curate Context (Add Knowledge)

Add project knowledge from the REPL:

/curate "Authentication uses JWT with 24h expiry and refresh tokens" @src/auth/jwt.ts

Or from the CLI:

brv curate -m "API rate limiting: 100 req/min per IP, enforced in middleware" -f src/middleware/ratelimit.ts

Curate operations are pending by default and require review.

Query Knowledge

Search the context tree:

/query How is authentication implemented?

Or:

brv query "What rate limiting rules are in place?"

Review Pending Changes

List pending curate operations:

brv review pending

Approve a specific operation:

brv review approve <operation-id>

Reject an operation:

brv review reject <operation-id>

View Curate History

brv curate view

Check Status

brv status

Shows project location, daemon status, active model, and sync state.

Version Control for Context

ByteRover supports Git-like version control for the context tree.

Initialize Version Control

brv vc init

This creates .brv/vc/ and sets up version tracking.

Stage and Commit Changes

brv vc add .                          # Stage all changes
brv vc commit -m "Add auth context"   # Commit staged changes

View Commit History

brv vc log

Branching

brv vc branch feature-auth            # Create a branch
brv vc checkout feature-auth          # Switch to it
brv vc branch                         # List branches

Merge Branches

brv vc checkout main
brv vc merge feature-auth

Push and Pull from Cloud

Push commits to ByteRover Cloud:

brv vc push

Pull commits from cloud:

brv vc pull

Clone a shared space repository:

brv vc clone <space-id>

Cloud Sync (Legacy)

The original push/pull commands create snapshots (not version-controlled):

brv push    # Snapshot context to cloud
brv pull    # Restore context from cloud

Prefer brv vc push / brv vc pull for version-controlled sync.

LLM Provider Configuration

List Available Providers

brv providers list

Supported: Anthropic, OpenAI, Google, Groq, Mistral, xAI, Cerebras, Cohere, DeepInfra, DeepSeek, OpenRouter, Perplexity, TogetherAI, Vercel, Minimax, Moonshot, GLM, OpenAI-Compatible, ByteRover.

Connect a Provider

brv providers connect

Follow the prompts to enter your API key (stored securely in .brv/config.json).

Or set environment variables:

export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...

Switch Active Provider

brv providers switch

List and Switch Models

brv model list
brv model switch

Real-World Usage Patterns

Pattern 1: Curate API Design Decisions

When you make a design decision, record it immediately:

brv curate -m "User endpoints: POST /users (create), GET /users/:id (read), PATCH /users/:id (update). All require JWT auth except public profile GET /users/:id/profile." -f src/routes/users.ts

Then query later:

brv query "What are the user endpoints and their auth requirements?"

Pattern 2: Document Complex Business Logic

brv curate -m "Order processing: validate inventory -> reserve stock -> charge payment -> fulfill. Rollback if payment fails. Stock reservation expires after 15 minutes." -f src/services/orders.ts

Pattern 3: Record Configuration Patterns

brv curate -m "Database: Postgres 14, connection pool max 20, timeout 30s. Migrations via Prisma. Replica for read-heavy queries in production." -f prisma/schema.prisma

Pattern 4: Share Context with Team

After curating knowledge locally:

brv vc add .
brv vc commit -m "Add order processing and DB config context"
brv vc push

Teammate pulls:

brv vc pull

They now have the same context tree.

Pattern 5: Query Before Implementing

Before writing new code, check existing patterns:

brv query "How do we handle error logging and monitoring?"

Use the answer to stay consistent with established patterns.

Worktrees and Knowledge Sources

What Are Worktrees?

A worktree is a subdirectory link to a parent project. It avoids creating nested .brv/ directories.

Use case: Monorepo with multiple packages, each needs ByteRover but should share one context tree.

Add a Worktree

From the project root:

brv worktree add ./packages/api

This creates .brv pointer file in ./packages/api that redirects to the parent .brv/.

List Worktrees

brv worktree list

Remove a Worktree

brv worktree remove ./packages/api

What Are Sources?

A source is a read-only reference to another project's knowledge.

Use case: Frontend app wants to query the backend's context tree without duplicating it.

Add a Source

brv source add ../backend-project

Now brv query searches both local and source knowledge.

List and Remove Sources

brv source list
brv source remove ../backend-project

MCP Server (Model Context Protocol)

ByteRover implements MCP for agent interop.

Start the MCP Server

brv mcp

This runs a JSON-RPC server that exposes ByteRover tools (curate, query, file ops, etc.) to MCP clients.

Configure MCP Client

Example claude_desktop_config.json:

{
  "mcpServers": {
    "byterover": {
      "command": "brv",
      "args": ["mcp"]
    }
  }
}

Claude Desktop can now call ByteRover tools directly.

Hub and Connectors

ByteRover Hub hosts reusable skills and bundles.

List Available Packages

brv hub list

Install a Package

brv hub install <package-name>

Add a Custom Registry

brv hub registry add my-registry https://registry.example.com

List and Remove Registries

brv hub registry list
brv hub registry remove my-registry

Connectors

Connectors extend ByteRover with integrations (e.g., GitHub, Slack).

brv connectors list
brv connectors install github-connector

Configuration

ByteRover stores config in .brv/config.json at the project root.

Typical Config Structure

{
  "projectId": "abc123",
  "spaceId": "space-xyz",
  "providers": {
    "anthropic": {
      "apiKey": "sk-ant-..."
    },
    "openai": {
      "apiKey": "sk-..."
    }
  },
  "activeProvider": "anthropic",
  "activeModel": "claude-3-5-sonnet-20241022",
  "reviewEnabled": true
}

Environment Variables

Override config with env vars:

export BRV_ANTHROPIC_API_KEY=sk-ant-...
export BRV_OPENAI_API_KEY=sk-...
export BRV_ACTIVE_PROVIDER=openai
export BRV_ACTIVE_MODEL=gpt-4o

Enable/Disable Review Workflow

By default, curate operations are pending. Disable review to auto-approve:

# Edit .brv/config.json
{
  "reviewEnabled": false
}

Or via env:

export BRV_REVIEW_ENABLED=false

Authentication to ByteRover Cloud

Login

Get an API key from app.byterover.dev/settings/keys, then:

brv login

Enter your API key when prompted.

Logout

brv logout

Troubleshooting

Daemon Not Running

If brv status shows "Daemon not running":

brv restart

API Key Not Recognized

Ensure the key is set:

brv providers list  # Check if provider is connected
brv providers connect  # Re-enter API key

Or verify env var:

echo $ANTHROPIC_API_KEY

Curate Operations Not Appearing

Check if review is enabled. If so, operations are pending:

brv review pending
brv review approve <operation-id>

Query Returns No Results

• Ensure you've curated knowledge first
• Check that files are indexed: brv status
• Verify active model supports embeddings

Version Control Conflicts

If brv vc pull fails due to conflicts:

brv vc status           # See conflicting files
# Manually resolve conflicts in .brv/vc/
brv vc add .
brv vc commit -m "Resolve merge conflicts"

Push Fails (Not Authenticated)

brv login
brv vc push

Worktree Not Recognized

Ensure .brv pointer file exists in the worktree directory and points to the correct parent:

cat packages/api/.brv
# Should contain: gitdir: ../../.brv

If missing, re-add:

brv worktree add ./packages/api

TypeScript API Usage (Advanced)

ByteRover is primarily a CLI, but you can use its internal APIs in TypeScript projects.

Install as Dependency

npm install byterover-cli

Example: Programmatic Curate

import { BrvClient } from 'byterover-cli';

const client = new BrvClient({
  projectRoot: process.cwd(),
});

await client.curate({
  message: "User service uses bcrypt for password hashing, cost factor 12",
  files: ['src/services/user.ts'],
});

console.log('Context curated');

Example: Programmatic Query

import { BrvClient } from 'byterover-cli';

const client = new BrvClient({
  projectRoot: process.cwd(),
});

const results = await client.query({
  query: "How is password hashing implemented?",
});

console.log(results.answer);

Example: Start REPL Programmatically

import { startRepl } from 'byterover-cli';

await startRepl({
  projectRoot: process.cwd(),
  provider: 'anthropic',
  model: 'claude-3-5-sonnet-20241022',
});

Best Practices

1. Curate Early and Often — Don't wait until the end of a feature. Curate decisions as you make them.
2. Use Specific File References — Always attach relevant files to curate operations for better context linking.
3. Enable Review in Team Settings — Prevent accidental or low-quality context pollution.
4. Commit Context Like Code — Use brv vc commit with meaningful messages for traceability.
5. Sync Regularly — Push and pull context changes daily in team environments.
6. Query Before Coding — Check existing patterns and decisions before implementing new features.
7. Use Worktrees for Monorepos — Avoid nested .brv/ directories by linking packages to a single root.
8. Add Sources for Cross-Project References — Link related projects to search their knowledge without duplication.
9. Leverage MCP for Agent Interop — Let other AI tools (Claude Desktop, etc.) access ByteRover's memory.
10. Document Configuration in Context — Curate environment setup, deployment steps, and config patterns so agents can reproduce them.

Common Workflows

Workflow 1: Solo Developer, Daily Usage

cd ~/my-project
brv                                   # Start REPL
# In REPL:
/curate "Redis caching: 1h TTL for user sessions, 5m for API responses" @src/cache.ts
/query "What are the caching rules?"
# Exit REPL
brv vc add .
brv vc commit -m "Add caching context"
brv vc push                           # Backup to cloud

Workflow 2: Team Collaboration

Developer A:

brv curate -m "API versioning: v1 in /api/v1, v2 in /api/v2. Deprecate v1 endpoints after 6 months." -f src/routes/index.ts
brv review approve $(brv review pending --json | jq -r '.[0].id')
brv vc add .
brv vc commit -m "Document API versioning policy"
brv vc push

Developer B:

brv vc pull
brv query "What is the API versioning policy?"
# Gets the answer immediately

Workflow 3: Onboarding New Team Members

New developer clones the repo and context:

git clone https://github.com/myteam/myproject.git
cd myproject
brv vc clone <space-id>               # Clone shared context
brv query "How do I set up the development environment?"
# ByteRover returns curated setup instructions

Workflow 4: Multi-Project Reference

Frontend project references backend context:

cd ~/frontend-app
brv source add ../backend-api
brv query "What are the available API endpoints?"
# Searches both frontend and backend context

Integration with AI Coding Agents

ByteRover works with 22+ AI coding agents. Here's how to integrate:

Cursor

Add to .cursorrules or workspace settings:

When working on this project, use ByteRover CLI (brv) to query and curate context.
Before implementing features, run: brv query "relevant question"
After making architectural decisions, run: brv curate -m "decision" -f <files>

Claude Code / Cline

In system prompt or project instructions:

This project uses ByteRover for persistent memory. Use `brv query` to search project knowledge before coding. Use `brv curate` to record important decisions and patterns.

Windsurf

Add to project docs or cascade rules:

# ByteRover Memory Layer
- Query context: `brv query "<question>"`
- Add context: `brv curate -m "<knowledge>" -f <files>`
- Sync with team: `brv vc push` after curating

MCP-Compatible Agents

Configure MCP client (e.g., Claude Desktop):

{
  "mcpServers": {
    "byterover": {
      "command": "brv",
      "args": ["mcp"]
    }
  }
}

Agent can now call ByteRover tools directly without shell commands.

Summary

ByteRover CLI is a memory layer for AI coding agents. It stores project knowledge in a version-controlled context tree, syncs to the cloud, and enables semantic search across team and project boundaries. Use brv curate to add knowledge, brv query to retrieve it, and brv vc push/pull to collaborate. Integrate with AI agents via shell commands or MCP for autonomous, context-aware coding workflows.