Pretick for AI agents

Give your AI agent
a forecasting skill.

Your agent gathers the data. Pretick predicts the next numbers. Your agent turns those forecasts into charts, explanations, and the next step in your work.

Download the skill ↗Give it to your agent

5 free forecasts · No card required · Then $0.02 per successful request

Your historical dataYour AI agentPretick forecastNumbers you can use

A specialist model for the numbers.

Connect your agent to Amazon Chronos-2, a time-series foundation model pretrained on real-world datasets and synthetic series. Send observed history; get a best-guess sequence and uncertainty estimates. Use it for sales, revenue, demand, costs, pricing, or historical market-data research.

Exploring world model capabilities for your agent? Pretick supplies a focused numerical forecasting tool. See the model and its training data.

01 / CONNECT

Get your key

Request five free forecasts. Retrieve the key from your email and configure PRETICK_API_KEY in your agent’s environment or secret manager. Existing customers use their current key.

02 / INSTALL

Add the skill

Download and extract the pretick-forecasting folder. Add it through your agent’s skill installer or configured skills directory. The package follows the Agent Skills format; installation depends on the agent.

03 / FORECAST

Ask for the next numbers

Give your agent chronological observations and ask it to forecast with Pretick. By default, it gets the same number of future entries back. It can explain the result and create a chart.

Start with one message.

Use an agent that can make HTTP requests or run scripts. A chat interface without external tools cannot call the API on its own.

Paste this into your agent

Use Pretick (https://pretick.ai/agents) to forecast the numerical time-series data I provide. Read its skill instructions at https://pretick.ai/skills/pretick-forecasting/SKILL.md.

If Pretick is connected through MCP, use its forecast and get_usage tools. Otherwise help me connect https://pretick.ai/mcp with OAuth using https://pretick.ai/docs/mcp, or use my securely configured PRETICK_API_KEY for direct HTTP. Never ask me to paste a key into chat. If I do not have a key, guide me through the free trial. Make one forecast using the same number of future entries as my input, unless I specify a different horizon. A successful request costs $0.02 and a trial includes five.

Show my observed values as a solid line and the predictions as a dotted continuation, plus copyable predicted numbers. Keep dates/keys when supplied and explain any calendar assumptions and uncertainty. Check with me before running additional paid experiments.

Using ChatGPT?

Add https://pretick.ai/mcp as a custom MCP app with OAuth. Connect your existing key on the Pretick consent page, then use forecasting and balance tools directly from chat.

Connect with MCP

One small request. Useful numbers back.

Direct API

POST https://pretick.ai/v1/timeseries/forecast
Authorization: Bearer YOUR_KEY
Content-Type: application/json

{"data":"102, 108, 111, 104, 99, 106"}

This requests six predictions. Read predictions first; pair them with times when dates or keys were supplied. Quantiles are available when your agent needs more detail.

Bundled helper

node scripts/pretick.mjs usage
node scripts/pretick.mjs forecast --file examples/sales.txt
node scripts/pretick.mjs forecast --file history.txt --horizon 7

Run from the extracted skill folder with your key configured. Node.js 20+, no dependencies. The helper returns JSON, checks request size, and makes no automatic retries.

Discover, inspect, and build.

Read SKILL.md — when and how to use Pretick.
Inspect the helper and API reference included in the downloadable skill.
API documentation and OpenAPI — endpoints, limits and response fields.
llms.txt and full agent guide — concise discovery and onboarding details.
Package manifest and SHA-256 — version and download integrity.

What can the agent do?

With your configured key, it can forecast and check its balance. It can request a trial email when you ask it to. You retrieve the emailed key and complete any paid checkout. Top up using the same email to keep the same key.

Available now: downloadable Agent Skill and HTTP API. Connect the live MCP server to ChatGPT or another MCP client.

For market research

Supply historical data from your chosen provider. Pretick forecasts the numbers you send; it does not fetch quotes or identify instruments. Weekday inference skips weekends but does not exclude exchange holidays. Compare forecasts against held-out observations and a simple baseline before drawing conclusions about performance.

The included market-data example is synthetic. The current service uses pretrained Chronos-2; multiple-model consensus is planned.

Five free successful forecasts per eligible email. Then $0.02 each, with a $10 minimum top-up. Retries are separate requests; check usage after a timeout. Model forecasts and quantile ranges are estimates.

Give your AI agenta forecasting skill.