velocity
Local-first workspace for Claude Code, Codex CLI, and Gemini CLI with sessions, analytics, workflows, and tools
claude mcp add --transport stdio optimilabs-velocity node server.js \ --env PORT="3000" \ --env BUN_LOG_LEVEL="info"
How to use
Velocity is a local-first command center that unifies Claude Code, Codex CLI, and Gemini CLI workflows into a single interactive interface. It provides three core loops: Run, where you execute tasks in local CLI sessions; Inspect, which surfaces usage metrics, latency, tool availability, and routing context; and Reuse, where proven patterns become Agents, Workflows, Skills, and Commands. With Velocity you can capture real work through Console and Sessions, transform repeated patterns into reusable automation, and visualize how data and commands flow through your routing graph. The built-in tooling supports building and running workflows, managing agents, and ordering complex command sequences with routing and analytics to help validate cost and correctness. You can regenerate actionable automation by turning successful sessions into reusable workflows and agents, then monitor their performance over time via the analytics surfaces.
What you can do in practice: create and modify workflows using the Workflow Builder, inspect session histories and transcripts, and route tasks through a graph to see how data flows between tools. Use Agents to encapsulate recurring tasks, Skills to define individual commands or prompts, and Hooks or Plugins to extend capability. Velocity also exposes a routing graph and analytics views to help you understand where time and cost are being spent, and to verify that automation remains correct as your setup evolves.
How to install
Prerequisites:
- Node.js or Bun-based environment (Velocity is built with Bun but can be run in compatible environments).
- Git installed to clone the repository.
Installation steps:
-
Clone the repository: git clone https://github.com/OptimiLabs/velocity.git cd velocity
-
Install dependencies (choose the recommended runner):
- If using Bun (recommended): bun install
- If using npm/yarn (alternative, depending on project setup): npm install
-
Start the development server:
- With Bun: bun dev
- If using node-based setup (alternative): node server.js
-
Open the app in your browser at http://localhost:3000
Notes:
- If you encounter PTY or terminal issues on macOS during setup, ensure proper permissions for any prebuilt binaries as described in the project's troubleshooting docs.
- If the port is in use, adjust the PORT environment variable or the server configuration accordingly.
Additional notes
Tips and common issues:
- Velocity runs locally and stores data on your machine by default; ensure you understand the local data implications if you enable remote sources.
- If Console PTY launch fails on macOS, run the suggested chmod commands to fix permissions for node-pty spawn-helper.
- When regenerating demos or sessions, you can specify a base URL and provider; ensure the provider is available locally or mocked for testing.
- If you modify routing or analytics, revalidate that your changes reflect correctly in the Routing Graph and analytics dashboards to avoid misinterpretation of cost or latency.
- Environment variables like PORT can be adjusted to fit your development environment; add any required API keys or tool credentials as environment variables if your workflows rely on external services.
Related MCP Servers
swiftlens
SwiftLens is a Model Context Protocol (MCP) server that provides deep, semantic-level analysis of Swift codebases to any AI models. By integrating directly with Apple's SourceKit-LSP, SwiftLens enables AI models to understand Swift code with compiler-grade accuracy.
neurolink
Universal AI Development Platform with MCP server integration, multi-provider support, and professional CLI. Build, test, and deploy AI applications with multiple ai providers.
roundtable
Zero-configuration MCP server that unifies multiple AI coding assistants (Codex, Claude Code, Cursor, Gemini) through intelligent auto-discovery and standardized interface
docmole
Dig through any documentation with AI - MCP server for Claude, Cursor, and other AI assistants
slack
Session-based Slack MCP for Claude and MCP clients: local-first workflows, secure-default HTTP.
devtap
Bridge build/dev process output to AI coding sessions via MCP — supports Claude Code, Codex, OpenCode, Gemini CLI, and aider