AI Integration & Skills

Ask your agent to investigate a company, read its filings, or check a financial calculation. Finance CLI provides the tools and source evidence; the skill helps your agent choose commands and handle the results.

You need an agent that can run terminal commands on your computer. The CLI supplies data retrieval and calculations, while your existing agent supplies the model and reasoning. Downloading the skill alone does not install the CLI.

Get started: Install the CLI → add the skill → ask your first question.

Download the Finance CLI skill

Ask your first research question

Once setup is complete, paste this into your agent:

Use Finance CLI to research Apple's latest annual filing. Explain its
business and three disclosed risks. Cite the filing date and source for
each finding. Retrieve more evidence when excerpts are truncated, and
report missing data or source errors.

The agent can discover the filing, read the relevant sections, and follow up with other commands. You should receive an explanation tied to filing evidence. Review the cited sources: the skill guides source handling, but does not guarantee that an agent’s interpretation is correct.

For another task, ask it to compare financial statement rows for companies you specify, or calculate growth from verified period values. See worked examples.

1. Install Finance CLI

Install or update the CLI first:

python -m pip install -U finresearch-cli

Verify the executable and local provider status:

finance --list
finance sources.status --output json

2. Download AI Docs

Agents should prefer the machine-readable files over scraping prose pages.

File	Use
llms.txt	Compact read order, routing rules, and trust reminders.
llms-full.txt	Full routing context and common playbooks.
tools.json	Canonical command schema, argument metadata, side effects, and citation fields.
openapi.json	Adapter contract for MCP/plugin/tool wrappers.
Agent Output Formats	Normalized records and compact renderers for LLM context compression.
finance-cli-skills.zip	Local skill package for agents that support skills.

3. Install The Skill

The Finance CLI skill is a compact routing guide for agents. It points them to the installed finance CLI, JSON output, tools.json schemas, and source-aware citation rules.

Copy this setup prompt into an agent that can manage local files:

Download https://tempestshaw.github.io/FinanceCLI/skills/finance-cli-skills.zip, extract it, copy the extracted skills/finance-cli folder into my local agent skills directory, verify that the copied folder contains SKILL.md, and then stop. Do not run pip install. Do not run finance commands during skill installation.

Package installation remains separate:

python -m pip install -U finresearch-cli

Manual Install

For Codex, install the skill folder into the local skills directory:

curl -L -o finance-cli-skills.zip https://tempestshaw.github.io/FinanceCLI/skills/finance-cli-skills.zip
rm -rf /tmp/finance-cli-skills
mkdir -p /tmp/finance-cli-skills
unzip -q finance-cli-skills.zip -d /tmp/finance-cli-skills
mkdir -p "${CODEX_HOME:-$HOME/.codex}/skills"
cp -R /tmp/finance-cli-skills/skills/finance-cli "${CODEX_HOME:-$HOME/.codex}/skills/"
test -f "${CODEX_HOME:-$HOME/.codex}/skills/finance-cli/SKILL.md"

For other agents that support local skills, copy the extracted skills/finance-cli folder into that agent’s skills directory.

Skill Overview

The skill is intentionally small:

File	Purpose
SKILL.md	Core routing rules for using finance <command> [arguments] --output json and the generated schema files.
docs/ROUTING.md	Namespace and command-family routing hints.
docs/PLAYBOOKS.md	Common multi-step workflows for filings, documents, market context, screens, and backtests.
docs/TRUST.md	Result-envelope handling, citation policy, credentials, and calculator boundaries.

The command schema truth remains tools.json. The skill links to schema files instead of duplicating them.

Usage Details

• Use finance --list to inspect commands in the installed version.
• Use finance sources.status --output json before assuming provider availability.
• Use tools.json for arguments, defaults, enums, output schemas, side effects, and citation fields.
• Preserve ok, data, error, and warnings from command output.
• Use compact record output only after deciding that the full JSON envelope is not needed for audit or replay.
• Keep finance calculations research-oriented and source-aware.