Act on your
business.
Enrich workflows with built-in tools for search, fetching, and code execution, or let models act on the systems you already run through the Model Context Protocol.
Ground every answer.
Add a tool type to any request and the model can search the web, read pages, and run code on its own. No infrastructure to stand up.
Let a model search the public web for fresh information. Two depth tiers cover anything from quick lookups to in-depth research.
type: 'web_search'Pull the readable content of one or more URLs. Pair with web_search to let the model both find and read pages.
type: 'web_fetch'Execute Python inside a fresh, isolated sandbox. Use it for math, data analysis, plotting, and anything that needs real execution.
type: 'code_interpreter'await client.chat.completions.create({
model: 'gpt-5-mini',
messages: [{ role: 'user', content: 'What changed in AI this week?' }],
tools: [
{ type: 'web_search' },
{ type: 'web_fetch' },
{ type: 'code_interpreter' },
],
})Let the model
run the code.
The code interpreter gives any tool-capable model a secure Python sandbox, so answers come from execution instead of estimation.
Real execution
The model writes Python and runs it, for math, data analysis, plotting, and anything a guess would get wrong.
Fresh and isolated
Every call gets a clean sandbox with nothing shared between calls, so one run can never leak into another.
Free during preview
The code interpreter costs nothing per call while in preview. You pay only for the model tokens.
File upload
for every model.
Files attached to a Chat Completions request are turned into text before they reach the model, so every model can read them, not only the ones with native file support.
Documents in, text out
Attach PDF, Word, PowerPoint, Excel, RTF, or plain text files and the content is extracted before the model sees it.
OCR for scanned pages
Image-only PDFs are read with OCR at no extra charge, so scanned contracts and receipts work like any other document.
Upload once, reuse
Upload through /v1/files and reference the file_id in any later request instead of sending the bytes again.
Call tools directly.
Some jobs do not need a model at all. Hit the tool endpoint and pay only for the pages or calls it used.
Run a web search and optionally scrape the resulting pages in one request.
POST /v1/tools/web_searchScrape a single URL with rich rendering options (markdown, HTML, screenshots, custom actions).
POST /v1/tools/web_scrapeDiscover every URL hosted on a domain. Useful for sitemaps, crawl planning, or scoping a downstream scrape job.
POST /v1/tools/web_mapAct on your business
over open standards.
Any remote MCP server
Add a tool of type mcp with a server URL. Aivene connects, discovers the tools, and runs every call the model makes.
One-click connectors
Connect Slack, Notion, GitHub, Figma, HubSpot, Stripe, Vercel, Intercom, PostHog, and more from the catalogue with OAuth.
Scoped to what you allow
Pass custom headers for auth and an allowed_tools list, so the model only sees the tools you choose.
Encrypted credentials
Connector tokens and headers are encrypted at rest and never returned to the client.
Same call as built-ins
MCP tools sit in the same tools array as web_search and your own functions. No new SDK to learn.
Your own functions too
Standard function calling works on every model that supports tools, mixed freely with MCP and built-ins.
await client.chat.completions.create({
model: 'gpt-5-mini',
messages: [{ role: 'user', content: 'Summarize open issues in our repo' }],
tools: [{
type: 'mcp',
server_label: 'github',
server_url: 'https://your-mcp-server.example.com/mcp',
allowed_tools: ['list_issues'],
}],
})Connected in
three steps.
- 01
Pick a connector
Open Integrations in the platform and choose from the catalogue, or add any MCP server by URL.
- 02
Authorize once
Sign in with OAuth or paste a token. Credentials are encrypted and stored against your account.
- 03
Call it from any model
Reference the server in your tools array. The model decides when to call it, and Aivene executes the call.
Questions about tools.
Design and manage
AI apps and agents.
Create, customize, and scale AI apps and agents on every major model. Get an API key and make your first call in minutes.
Pay as you go. No monthly minimum.
