Fetch Agents
About
AI-powered code analysis and context optimization tools. Slim large code files to relevant sections (save 60-80% tokens), compress text, analyze conversations for noise, and search past chat knowledge bases. Supports x402 micropayments on Base.
Explore
- Reduce code files to only relevant sections with slim_code
- List and extract LAYER sections from code files
- Compress long text while preserving key information
- Analyze conversations for noise and wasted tokens
- Index chat exports into a searchable knowledge base
- Search indexed knowledge bases for past discussion insights
- Pay-per-use via x402 micropayments on Base
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
Fetch AgentsCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
- Python 3.10+
- mcp and httpx packages
- Internet connection (tools call cloud APIs)
Claude Code (~/.claude/.mcp.json):
{
"mcpServers": {
"fetch-agents": {
"command": "python",
"args": ["mcp_server.py"],
"cwd": "/path/to/fetch-agents-mcp",
"env": {
"FETCH_API_KEY": "fak_your_key_here"
}
}
}
}
Cline (VS Code settings):
{
"cline.mcpServers": {
"fetch-agents": {
"command": "python",
"args": ["mcp_server.py"],
"cwd": "/path/to/fetch-agents-mcp",
"env": {
"FETCH_API_KEY": "fak_your_key_here"
}
}
}
}
pip install mcp httpx
Once installed, your AI assistant can use the tools naturally:
> "Use slim_code to get only the authentication-related code from this file"
> "Compress this error log to the key information"
> "Search my teleported knowledge base for how we handled rate limiting"
| Tool | Description | Use When |
|------|-------------|----------|
| slim_code | Reduce a code file to only sections relevant to your task | Reading large files, saving tokens |
| map_layers | List all LAYER sections in a code file | Exploring unfamiliar code structure |
| extract_layer | Extract one specific named LAYER section | Focused reading of a single component |
| compress_text | Compress long text while preserving key information | Summarizing docs, logs, verbose output |
| analyze_context | Find noise and wasted tokens in conversations | Long conversations hitting context limits |
| teleport_chat | Index a chat export into searchable knowledge base | Reusing insights from past conversations |
| search_knowledge | Search indexed knowledge bases | Finding relevant past discussions |
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"fetch agents": {
"fetch-agents": {
"command": "python",
"args": [
"mcp_server.py"
],
"cwd": "/path/to/fetch-agents-mcp"
}
}
}
}McpServers
{
"fetch-agents": {
"command": "python",
"args": [
"mcp_server.py"
],
"cwd": "/path/to/fetch-agents-mcp"
}
}MCP server providing AI-powered code analysis, context compression, and knowledge search tools for Claude Code, Cline, and other MCP-compatible AI assistants.
Tools
| Tool | Description | Use When |
|------|-------------|----------|
| slim_code | Reduce a code file to only sections relevant to your task | Reading large files, saving tokens |
| map_layers | List all LAYER sections in a code file | Exploring unfamiliar code structure |
| extract_layer | Extract one specific named LAYER section | Focused reading of a single component |
| compress_text | Compress long text while preserving key information | Summarizing docs, logs, verbose output |
| analyze_context | Find noise and wasted tokens in conversations | Long conversations hitting context limits |
| teleport_chat | Index a chat export into searchable knowledge base | Reusing insights from past conversations |
| search_knowledge | Search indexed knowledge bases | Finding relevant past discussions |
Quick Start
1. Get a free API key
curl -X POST https://agentslab.duckdns.org/code/register \
-H "Content-Type: application/json" \
-d '{"name": "my-project"}'
50 free requests/day included.
2. Add to your MCP config
Claude Code (~/.claude/.mcp.json):
{
"mcpServers": {
"fetch-agents": {
"command": "python",
"args": ["mcp_server.py"],
"cwd": "/path/to/fetch-agents-mcp",
"env": {
"FETCH_API_KEY": "fak_your_key_here"
}
}
}
}
Cline (VS Code settings):
{
"cline.mcpServers": {
"fetch-agents": {
"command": "python",
"args": ["mcp_server.py"],
"cwd": "/path/to/fetch-agents-mcp",
"env": {
"FETCH_API_KEY": "fak_your_key_here"
}
}
}
}
3. Install dependencies
pip install mcp httpx
x402 Payments (No API Key Needed)
All endpoints support x402 micropayments on Base (Coinbase L2). Any AI agent with a USDC wallet can pay per-request without registration:
POST /code/slim -> 402 Payment Required + payment-required header
-> Agent signs USDC transfer -> retries with payment header -> gets response
Prices: $0.10 - $1.00 per request depending on tool.
How It Works
The MCP server connects to cloud-hosted agents at agentslab.duckdns.org via HTTPS. Your code is processed server-side using LLM-powered analysis pipelines. No local GPU needed.
Your IDE -> MCP Server -> HTTPS -> Cloud Agents -> Response
Example Usage
Once installed, your AI assistant can use the tools naturally:
> "Use slim_code to get only the authentication-related code from this file"
> "Compress this error log to the key information"
> "Search my teleported knowledge base for how we handled rate limiting"
Pricing
| Tier | Price | Included |
|------|-------|----------|
| Free | $0 | 50 requests/day |
| Pay-per-use (x402) | $0.10-1.00 | Per request, USDC on Base |
| Prepaid balance | Top up via crypto | BTC, ETH, USDC, + 10 more |
| Pro subscription | $15/mo | 300 requests/mo |
Requirements
- Python 3.10+
- mcp and httpx packages
- Internet connection (tools call cloud APIs)
License
MIT
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
