Browser Agent Overview
Your AI controls a real browser
Crawlio's browser agent gives AI applications full control of a web browser. There are two modes. Both use the same tool interface. Your AI does not need to know which one is active.
| Mode | When to use | Tools | Install |
|---|---|---|---|
| Live browser (Chrome extension) | You have Chrome open. AI controls your tabs directly via CDP. | 96 tools | Chrome Web Store |
| Headless browser (background agent) | No Chrome needed. Runs in the background. Converts, intercepts, automates. | ~199 tools across 5 tiers | Bundled with Crawlio or npx |
How routing works
The aggregator handles routing automatically:
- If a Chrome tab is connected, browser tools route to the extension.
- If no tab is connected, they route to the headless agent.
- Crawlio app tools (crawl engine, export, license) always route to the app.
Five meta-tools unify everything into a single namespace:
| Meta-tool | Purpose |
|---|---|
crawlio_discover |
List available tools across all pillars |
crawlio_call |
Route a tool call to the correct pillar |
crawlio_do |
High-level task execution with automatic pillar selection |
crawlio_cortex |
Intelligence queries across pillar data |
crawlio_consult |
Multi-pillar consultation for complex tasks |
Key capabilities
- Navigate pages, click, type, scroll, take screenshots
- Capture network requests, console logs, DOM snapshots
- Detect JavaScript frameworks at runtime (React, Vue, Angular, Svelte, Next.js, and more)
- Convert pages to PDF, screenshots, and other formats
- Intercept and modify network traffic in flight
- Feed all captured data back to Crawlio for enrichment
SessionVault
Both modes can use stored auth sessions. Log in once through the vault, and browser tools handle authenticated pages automatically.
The vault is managed by the Crawlio app (the sole decryption authority). Five vault tools control session lifecycle:
| Tool | Purpose |
|---|---|
vault_request_login |
Request interactive login for a domain |
vault_list_domains |
List domains with stored sessions |
vault_get_session |
Retrieve a decrypted session |
vault_mark_expired |
Mark a session as expired |
vault_delete |
Delete a stored session |
The Chrome extension uses native Chrome session storage directly and does not need the vault.
Crawlio app integration
If Crawlio.app is running locally, the browser agent can:
- Start crawls via
extract_siteor the Crawlio HTTP API - Send enrichment data: framework detection, network requests, console logs, DOM snapshots
- Read crawl state: status, downloads, enrichment data
- Export results: trigger exports in any format
Network request URLs discovered by the browser are automatically offered to the crawl engine's frontier.
Getting started
Next steps
- Browser Tools Reference: all 96 Chrome extension tools
- JIT Context Runtime: polymorphic context and the agentic REPL
- MCP Overview: Crawlio's MCP server for crawl control