Tools your models can actually use
Six built-in tools are included with every account. Enable them per conversation and the model starts calling them automatically — no API keys, no configuration files, no extra setup.
Built-in toolbox
Six tools, zero setup
Every tool is sandboxed, streamed back inline, and visible in the thread. Toggle any of them on or off per conversation without touching a config file.
Web search
The model queries the live web and surfaces relevant results directly inside the conversation — ideal for current events, documentation lookups, and real-time research.
- Real-time results from the open web
- Cites sources with links in the reply
- Configurable per-chat search depth
HTTP request
Fetch any public URL or make authenticated REST calls from within the conversation. Useful for checking live data, probing APIs, and building ad-hoc integrations on the fly.
- GET, POST, PUT, and DELETE support
- Custom headers and bearer-token auth
- Response body parsed and summarised inline
Calculator
Evaluate mathematical expressions accurately without relying on the model's arithmetic. Handles algebraic formulas, unit conversions, and common statistical functions.
- Exact arithmetic — no hallucinated results
- Supports units, fractions, and exponents
- Step-by-step expression breakdowns on request
Date & time
Query the current date and time in any timezone, calculate durations between two dates, and format timestamps — anchoring every conversation to real-world time.
- Current time in any IANA timezone
- Duration and age calculations
- ISO 8601 and locale-aware formatting
Document reader
Upload PDFs, Word documents, spreadsheets, and plain-text files for the model to read and analyse inline — no manual copy-pasting or pre-processing required.
- PDF, DOCX, XLSX, and TXT file support
- Intelligent chunking for large documents
- References sections and page numbers
Code interpreter
Execute Python, JavaScript, or shell snippets inside an isolated sandbox. Perfect for data processing, chart generation, and validating logic without leaving the chat.
- Secure sandboxed execution — no host access
- Python, Node.js, and Bash runtimes included
- stdout, stderr, and artefacts streamed back
How it works
Automatic tool-calling
You don't pick which tool runs — the model does. Based on the conversation context it decides whether to search the web, run code, or crunch a number. Results stream back inline before the reply continues.
At each turn the model receives the list of enabled tools alongside your message. When a tool call is warranted it dispatches the request to a secure server-side runner, validates inputs and outputs, and streams the result — a search snippet, a computed answer, a file excerpt — directly into the thread.
Every call and its result are fully visible in the conversation so you always know what the model fetched, computed, or executed. Multi-step tool chains — search, then summarise, then calculate — execute transparently in a single turn.
Need something custom?
The built-in tools cover the most common cases, but every team is different. Dynamic tools let you connect any HTTP API as a model-callable tool — define a name, a URL, and a JSON parameter schema — no code required.Give your models real capabilities
Enable built-in tools in seconds. Extend further with dynamic HTTP tools whenever you need to connect your own APIs.