FN2 for Claude Code | Research agents from your terminal
For developers

Use FN2 from Claude Code

Start, schedule, and read FN2 research agents without leaving your terminal. One plugin, no API key to copy, and a scoped credential that never touches chat.

$claude plugin marketplace add fn2ai/fn2-skill && claude plugin install fn2@fn2

Two ways to connect

FN2 reaches Claude Code through two front doors. Use the skill to drive agents from the terminal, or add the MCP connector to expose FN2 research data to any MCP client.

The fn2 skill

A Claude Code plugin that drives your agents from the terminal. Create, schedule, run, pause, and read results in natural language, with no API key to manage.

claude plugin install fn2@fn2

The MCP connector

Add the FN2 MCP server to any MCP-compatible client and expose FN2 research data directly to the model. Read-only market intelligence, no agent management.

https://mcp.fn2.ai/mcp

Set up in three steps

No pip install, no build step, no secret to paste. The skill is one dependency-free Python 3 file tree.

1

Add the marketplace and install

Add the fn2-skill repository as a Claude Code marketplace, then install the plugin.

$ claude plugin marketplace add fn2ai/fn2-skill
$ claude plugin install fn2@fn2
2

Connect your account

Run the login command. The terminal prints a short code; open fn2.ai/start, sign in, type the code, and approve. The CLI receives a scoped, revocable device credential stored in your keychain. You never see or copy a secret.

$ fn2 auth login
# Open https://fn2.ai/start and enter code: XXXX-XXXX
$ fn2 whoami          # confirms plan + quota
3

Ask your assistant

Talk to Claude Code in natural language. It calls the skill to start agents, check results, and put them on a schedule.

you › start an agent that tracks unusual options flow
      in my watchlist hourly
  ↳ ✓ agent created, runs hourly, results emailed

What you can do

Everything the fn2 web app does, driven from a prompt. Schedule in plain English, read cited results, and manage a fleet of background researchers.

Run agents on a schedule

Start an agent that runs once or on a cadence. Say "before market open" or "weekdays at 8am" and the skill resolves it against the exchange calendar, skipping holidays and half-days.

Read cited research

Pull live prices, search earnings transcripts, read SEC filings, and query FRED macro series. Every answer cites its sources, with charts and tables where they help.

18 tools at your disposal

The skill and MCP connector expose the same fn2 surface: five read-only research tools and thirteen agent and model tools.

Research (read-only)

5 tools
  • search_transcripts
  • recall
  • get_stock_price
  • get_fred_series
  • get_company_financials

Agents and models

13 tools
  • list_agents
  • get_agent
  • start_agent
  • update_agent
  • pause_agent
  • resume_agent
  • delete_agent
  • run_agent
  • list_runs
  • get_run
  • get_usage
  • list_models
  • run_chat

Weekly research allowance

Every plan gets a weekly research-call allowance that resets Monday 00:00 UTC. Research calls spend no model tokens, so this is separate from your monthly chat quota.

Plan Weekly research calls
Free25
Pro trial100
Pro500
Ultra2,000
Max5,000

Allowance resets Monday 00:00 UTC. Agent runs and chat turns draw from your monthly token quota; research tool calls do not.

Start building with FN2 and Claude Code

Create a free account, install the skill, and have your first scheduled agent running in minutes. No credit card required.