About context-bridge
ContextBridge is an MCP server for compressing AI conversation transcripts into structured handoffs and preparing them for another AI platform. It is for people moving ongoing work between AI conversations without repeating context.
Key features
- 01Compresses transcripts into structured context
- 02Preserves goals, decisions, and open threads
- 03Formats handoffs for Claude, ChatGPT, and Gemini
- 04Builds destination-ready resume packages
Use cases
- 01Hand off ongoing work to another AI
- 02Carry decisions and constraints into a new conversation
MCP support
Exposes `compress_context`, `format_for_target`, and `resume_context` to MCP clients. The server runs over stdio and supports provider credentials or a configured hosted lane.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @contextbridge_ai/mcpOPENAI_API_KEYSecretOptional BYOK API key for the OpenAI Responses or chat-completions lane.OPENAI_MODELOptional model override for the OpenAI BYOK lane.CONTEXTBRIDGE_HOSTED_URLOptional hosted inference endpoint for the no-key lane.CONTEXTBRIDGE_HOSTED_API_KEYSecretOptional auth key for the hosted inference endpoint.CONTEXTBRIDGE_HOSTED_MODELOptional hosted model identifier override.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
4Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| name | string | Yes | The name to greet |
Details
From the project READMEContextBridge
ContextBridge is a context portability tool for AI conversations. This package publishes the MCP server surface: it compresses an in-progress conversation into a structured context block that preserves the user's goal, confirmed decisions, current state, open threads, artifacts, constraints, recent pivots, and suggested next step, then formats that handoff so another AI can continue without forcing the user to restate everything.
The published npm artifact is the stdio MCP server for Claude Desktop, agent runtimes, and other MCP-compatible hosts. The browser extension runtime is developed in the same repository, but it is not part of this MCP Registry publishing flow.
Published Links
- npm package: @contextbridge_ai/mcp
- MCP Registry search: io.github.prateekg7/context-bridge
- MCP Registry API: latest version JSON
- Source repository: prateekg7/Context-Bridge
Install In Claude Desktop
Add this to your Claude Desktop MCP config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"contextbridge": {
"command": "npx",
"args": ["-y", "@contextbridge_ai/mcp"],
"env": {
"GROQ_API_KEY": "your_groq_api_key_here"
}
}
}
}
Or with OpenAI:
{
"mcpServers": {
"contextbridge": {
"command": "npx",
"args": ["-y", "@contextbridge_ai/mcp"],
"env": {
"OPENAI_API_KEY": "your_openai_api_key_here",
"OPENAI_MODEL": "gpt-4.1-mini"
}
}
}
}
For a hosted no-key lane instead of BYOK:
{
"mcpServers": {
"contextbridge": {
"command": "npx",
"args": ["-y", "@contextbridge_ai/mcp"],
"env": {
"CONTEXTBRIDGE_HOSTED_URL": "https://your-hosted-endpoint.example/v1",
"CONTEXTBRIDGE_HOSTED_MODEL": "mistralai/Mistral-7B-Instruct-v0.2"
}
}
}
}
Available MCP Tools
compress_context
Compresses a raw transcript into a structured context block.
Parameters:
transcript(string, required): the raw conversation transcript to compress.provider("openai" | "openai-chat" | "google" | "groq" | "custom" | "hosted", optional): provider override.model(string, optional): model override for the selected lane.apiKey(string, optional): BYOK API key override.baseUrl(string, optional): API base URL override.hostedApiKey(string, optional): optional auth key for the hosted lane.
format_for_target
Formats a compressed context block for a destination platform.
Parameters:
context_block(object, required): a validated ContextBridge compressed context block.target("claude" | "chatgpt" | "gemini", required): destination AI platform.
resume_context
Builds a destination-ready landing package from a compressed context block.
Parameters:
context_block(object, required): a validated ContextBridge compressed context block.target("claude" | "chatgpt" | "gemini", required): destination AI platform.
BYOK Setup
ContextBridge auto-selects the OpenAI Responses lane when a BYOK key is present.
Environment variables:
OPENAI_API_KEY=your_openai_api_key_here
OPENAI_MODEL=gpt-4.1-mini
OPENAI_BASE_URL=https://api.openai.com/v1
Additional supported provider overrides:
GOOGLE_API_KEY=your_google_api_key_here
GOOGLE_MODEL=gemini-2.5-pro
GOOGLE_BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai
GROQ_API_KEY=your_groq_api_key_here
GROQ_MODEL=llama-3.3-70b-versatile
GROQ_BASE_URL=https://api.groq.com/openai/v1
CONTEXTBRIDGE_API_KEY=your_custom_api_key_here
CONTEXTBRIDGE_MODEL=your_custom_model_here
CONTEXTBRIDGE_BASE_URL=https://your-endpoint.example/v1
CONTEXTBRIDGE_PROVIDER=custom
Hosted Lane Setup
The hosted lane is the no-key path for end users who do not bring their own provider credentials.
Environment variables:
CONTEXTBRIDGE_HOSTED_URL=https://your-hosted-endpoint.example/v1
CONTEXTBRIDGE_HOSTED_MODEL=mistralai/Mistral-7B-Instruct-v0.2
CONTEXTBRIDGE_HOSTED_API_KEY=
Selection priority:
- Explicit
apiKeyoption passed into compression uses the OpenAI Responses BYOK lane. OPENAI_API_KEYuses the OpenAI Responses BYOK lane.CONTEXTBRIDGE_HOSTED_URLuses the hosted lane.
Contributing
Contributions are welcome. The highest-leverage areas are extraction quality evaluation, browser collector resilience, destination formatting quality, hosted-lane infrastructure, and real-world handoff testing across platforms.
License
MIT. See LICENSE.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.