Model Context Protocol

Your AI agents,
inside Claude.

Install the Daily AI Agent OS MCP server and use all 9 agents as native tools inside Claude Desktop, Claude Code, or any MCP client. Local-first. No cloud. One command to install.

A standard protocol for AI tool use

The Model Context Protocol (MCP) is an open standard that lets AI assistants like Claude call external tools and data sources — without custom integrations for every use case.

By running this MCP server, you make the entire Daily AI Agent OS available as first-class tools inside Claude. Ask Claude to research a topic, generate a content draft, or run a backtest — and it calls the real agents running locally on your machine.

All computation stays local. No data leaves your machine.

Claude Desktop
or Claude Code
Model Context Protocol (stdio)
daily-ai-agent-os MCP server
Research agents
Trading agents
Content agents
Local LLM (Ollama / LM Studio) + Local data

Everything in one server

Each tool maps directly to a production agent in the OS.

🔍
research
Research

Single-pass research on any topic. Decomposes questions, researches sub-topics, synthesizes findings with the local LLM.

research("SQLite vs DuckDB for local analytics")
🧠
deep_research
Research

Multi-pass research orchestrator. Plans strategy, runs iterative investigation passes, queries the local knowledge base.

deep_research("Best RAG approach for exocortex?", depth=2)
✍️
generate_content
Content

Generate blog posts, X threads, or newsletters. Grounded by local research. Runs a quality gate before returning.

generate_content("Why local AI matters", format="thread")
quality_check
Content

Fact-check any text. Detects hallucinated metrics, fabricated performance numbers, and implausible round-number claims.

quality_check("Our system hit a 94% health score...")
📈
trading_signals
Trading · Paper only

Get BUY/SELL/HOLD signals for crypto. Live prices from CoinGecko, strategies run in dry-run mode. No real money.

trading_signals("BTC,ETH,SOL")
⚗️
backtest_strategy
Trading · Paper only

Run momentum, mean-reversion, trend-following, or combined strategy backtests. Returns Sharpe, drawdown, win rate, and more.

backtest_strategy("BTC,ETH", strategy="momentum")
📊
market_data
Trading

Historical OHLCV data for crypto or stocks. Pulls from local Parquet files, falls back to CoinGecko or yfinance.

market_data("BTC", days=90)
📚
knowledge_search
Knowledge

Search the local knowledge index. Finds prior research reports, decisions, and facts stored in the exocortex.

knowledge_search("ollama inference optimization")
🩺
system_health
System

Full system health report. Checks inference provider, exocortex integrity, agent modules, disk space, and git state.

system_health()

Up in three steps

Works with Claude Desktop, Claude Code, and any MCP-compatible client.

01

Clone the repo

Get the Daily AI Agent OS on your local machine. Requires Python 3.11+.

02

Run the installer

One script installs the MCP SDK and adds the server to your Claude Desktop config.

03

Restart Claude

Restart Claude Desktop (or reload Claude Code). The tools appear automatically.

Quick Install (Mac)
# 1. Clone and enter the repo git clone https://github.com/dailyaiagents/daily-ai-agent-os.git cd daily-ai-agent-os # 2. Run the installer bash mcp_server/install.sh # 3. Restart Claude Desktop — tools are ready
Manual — claude_desktop_config.json
{ "mcpServers": { "daily-ai-agent-os": { "command": "python", "args": ["-m", "mcp_server"], "cwd": "/path/to/daily-ai-agent-os" } } }
Claude Code CLI
claude mcp add daily-ai-agent-os python -- -m mcp_server

Local AI, inside your AI.

Everything the OS can do — available anywhere Claude runs.