Skip to main content

Using with AI Agents

The TraderBro CLI is designed to be driven by AI assistants. Every command supports --json for machine-readable output. The flag tables and JSON schemas in this documentation are structured so AI models can parse them reliably.

Setup for Claude Code​

Install the SKILL.md into your Claude Code session to give Claude full knowledge of every command:

# In your project, point Claude Code to the SKILL.md from the CLI repo
# Or set TRADERBRO_API_KEY in your environment and run:
traderbro whoami --json

Once Claude Code knows the CLI is available, you can ask it directly:

"Use the traderbro CLI to find the top 5 analysts by 3-month return with at least 20 predictions, then get their sector edge in Technology."

Claude will chain these commands automatically:

traderbro analyst list --period 3m --sort return --min-predictions 20 --limit 5 --json
traderbro analyst sector-edge <slug> --sector Technology --period 3m --json

Copy-pasteable prompts​

These prompts work with Claude, GPT-4, and similar AI assistants that have CLI access.

Find top analysts:

"Run traderbro analyst list --sort accuracy --min-predictions 20 --limit 10 --json and tell me which analysts have the best combination of accuracy and return."

Sector-specific research:

"Find analysts with strong Technology sector edge using traderbro analyst sector-edge <slug> --period 3m --group-by industry --json. Check the top 3 analysts by accuracy."

Symbol sentiment check:

"Get all NVDA predictions from the last 7 days using traderbro prediction list --symbol NVDA --since <date> --json and summarise the bull/bear sentiment."

Period comparison:

"Compare the top 10 analysts by 1-month return vs 3-month return. Use traderbro analyst list --period 1m --sort return --limit 10 --json and traderbro analyst list --period 3m --sort return --limit 10 --json."

Always use --json with AI agents​

Table output is for human reading. JSON gives the AI the full data structure it needs to reason over results:

# Bad for agents — truncated table, no numeric precision
traderbro analyst list --limit 10

# Good for agents — full structured data
traderbro analyst list --limit 10 --json

Pagination in agentic loops​

For large result sets, use --limit and --page to paginate:

traderbro prediction list --symbol NVDA --json --limit 50 --page 1
traderbro prediction list --symbol NVDA --json --limit 50 --page 2

The JSON response includes count and next fields. Ask the AI to stop paginating when next is null.

Machine-readable schema​

To give an AI a complete map of all commands and flags without reading the docs:

traderbro describe --json

This outputs a JSON schema of every command, subcommand, flag, type, and default value. AI agents can fetch this once and use it to construct valid commands autonomously.