Tools
Connected Tools give your AI agents eyes and hands, enabling them to retrieve live information from the web, interact with real-world APIs, resolve geographic locations, and run calculations.
Without tools, large language models are limited to the static snapshot of their pre-training data. With tools enabled, agents can dynamically look up real-time facts, verify sources, inspect live URLs, and perform deterministic computation.
Available Native Tools
AIsh.li includes a suite of native tools that can be enabled on any agent with a single click:
| Tool | ID | Description | Guide |
|---|---|---|---|
| Google Search | search-google |
Real-time web search grounding with live source citations. | Google Search Guide |
| Web Fetch | http |
Scrape public web pages, read documentation, and inspect live APIs. | Web Fetch Guide |
| Google Maps | geo-google-maps |
Geocode addresses, resolve GPS coordinates, and locate places. | Google Maps Guide |
| Code Execution | code-execution |
Execute sandboxed Python code for exact mathematical calculations. | Code Execution Guide |
How Tool Calling Works
When a user asks a question that requires external information or calculation:
sequenceDiagram
autonumber
actor User
participant Model as AI Model
participant Runtime as Tool Runtime
participant Search as Google Search API
User->>Model: "What was the score of yesterday's Champions League match?"
Note over Model: Recognizes need for real-time live data
Model->>Runtime: Tool Call: search-google(query="Champions League score yesterday")
Runtime->>Search: Execute web query
Search-->>Runtime: Structured search results & snippets
Runtime-->>Model: Grounded tool response
Model-->>User: Synthesized answer with citations
Enabling Tools on an Agent
You can enable or disable tools for any agent in two ways:
Visual Builder:
- Go to the Agents page and select an agent to edit (or Create a New Agent).
- Scroll down to the Tools section.
- Toggle the checkboxes for HTTP/URL π, Google Search π, Google Maps πΊοΈ, or Code Execution π».
- Click Save Agent.
Declarative YAML:
- Click the Code tab on any agent page.
- Add the desired tool identifiers to the
toolsarray:
id: my-account/research-assistant
instruction: You are a thorough researcher. Always verify facts using search.
model: gemini-3.8-flash
tools:
- search-google
- http
Model Context Protocol (MCP)
In addition to our native built-in tools, AIsh.li is built with support for the open Model Context Protocol (MCP), allowing enterprise agents to connect to proprietary databases, internal APIs, and custom enterprise tools.
Next Steps
Explore the detailed guides for each native tool:
- π Google Search Grounding: Real-time search and fact verification.
- π Web Fetch Tool: Retrieving and reading live URLs.
- πΊοΈ Google Maps & Geolocation: Geographic resolution and coordinates.
- π» Code Execution: Running sandboxed Python computation.