Open source

Agent Kits

Research tools for your coding agent.

Start with a Python demo, inspect the output, and adapt it to your research. Setup guides and shared tools are included.

Get the starter prompt Explore the workflows

Free and open source. Offline examples need no account or API key.

Synthetic SPY gamma replay. Strike 100 moves from −100 to −120 to −70, then a gap, then −50 and 0. Strike 105 also has gaps.

Sample data

Exposure by strike

Signed exposure for fixed strikes 100 and 105. Missing strikes and changed expirations stay visible.

Chart drawn from the saved report.

Read original output Opens in a new tab. Save graphic

Setup

Start with your coding agent.

Use a coding agent that can run local commands. It can check your setup, run the node-tracker demo and help you choose what to build next.

Git and Python 3.11+ are required. The first demo needs no Skylit key, model API key or Python packages. Your coding agent may have its own costs.

Guided setup GitHub. Opens in a new tab.
Paste into your coding agent
Get me started with https://github.com/SkylitAI/skylit-agent-kit. In a repository-scoped session, read AGENTS.md and docs/start-here.md before running commands. Handle setup, run the offline node-tracker demo, and show me the chart. Use docs/capabilities.md to help me choose and build my next workflow. Keep private vaults out and never request credentials in chat. Use live calls only within my authorized budget; publish only with my authorization.

Workflows

Choose an example.

Each example runs offline with sample data and uses no API credits. Change an input to see how the report changes.

Fixed-strike replay

Follow the same strikes through signed exposure changes. See the gaps instead of filling them in.

use-case node-tracker

Prices beside levels

Put dated OHLCV bars beside separately timed exposure levels. Keep their source times distinct.

use-case price-levels

Flow investigation

Inspect a returned trade sample alongside strike rollups for an explicit time window.

use-case flow-investigator

Volatility context

Review returned volatility fields with freshness and coverage notes.

use-case volatility-context
Commands and workflow guides GitHub. Opens in a new tab.Explore Lab experiments

Included

Tools you can build on.

Agent Kits is built around the Skylit Agent Kit repository: maintained examples, shared helpers and agent setup guides. Run Python directly or have your agent help.

Browse the repository GitHub. Opens in a new tab.
77 endpoint previews
Synthetic request and response examples across the public API. Inspect the data format before connecting an account.
Watchlist reports
Plan a GEX, VEX and recent-flow report first. Review requests and documented credits before an explicit live run.
Codex and Claude guides
Setup guidance for local workflows and optional MCP connections. Host certification is still pending.
Reports with context
Keep source timestamps, missing observations and coverage limits beside the result.

Live data

Review requests before connecting.

The offline examples are ready to explore. Live paths are implemented and tested with synthetic responses; authenticated service certification remains pending.

  1. 1

    Check your account access

    Skylit service access and API credits depend on your account. Open the Developer page.

  2. 2

    Review the request budget

    Start with a dry run. Live mode is explicit; the runner checks the plan against its request and credit caps. Direct MCP calls do not inherit those caps.

  3. 3

    Keep your key local

    Use the secure environment or hidden terminal prompt. Never paste credentials into agent chat.

Read the public API docs

Common questions

Do I need a Skylit subscription to try the demo?

The offline examples do not call Skylit services and need no Skylit account or API key. The repository is public. Agent subscriptions may have their own costs.

Is this a hosted agent or an app inside Skylit?

It is a repository of local tools, examples and guides you can run and adapt. For Skylit’s in-app AI analyst, see Talon.

Can I use MCP with my agent?

Optional MCP setup is covered in the agent guides. Direct MCP calls have their own access and costs and do not inherit the local runner’s limits. Start with the public API documentation and the repository’s host-specific guide.

Where should I share a new workflow?

Start an experiment in Agent Lab. Reusable work can graduate into the maintained Kit through review. Code licensing does not include service access or third-party data rights.

Get started

Try Agent Kit.

Explore Kit on GitHub Explore Agent Lab

For research and education. These workflows do not place trades. Live services and agent hosts have not been independently verified.