Agent Mode

Let the AI autonomously search, fetch URLs, read files, and call tools to answer complex legal questions

Agent Mode gives the AI autonomous control over which tools to use and when. Instead of following a fixed retrieval pipeline, the AI decides what to search, which URLs to read, and what tools to call - iterating up to 5 rounds until it has a complete answer.

How It Works

In regular chat, your query goes through a fixed RAG pipeline: embed, retrieve, rerank, generate. In Agent Mode, the AI receives a set of tools and autonomously decides the research strategy:

1. You ask a question
2. The AI analyzes it and calls relevant tools (search documents, web search, fetch a URL, etc.)
3. It reads the results, identifies gaps, and calls more tools if needed
4. After up to 5 iterations, it synthesizes a final answer with sources

When Agent Mode is enabled, Research and Deep Mode are paused. The agent takes over all search autonomously.

Enable Agent Mode

Click the Agent button in the chat toolbar. A toggle turns it on/off, and a dropdown lets you configure individual tools.

Agent Mode is saved per matter - enable it once and it stays on for that workspace.

Available Tools

Agent Mode gives the AI access to these tools automatically:

Tool	What It Does	Cost
Web Search	Searches the web with jurisdiction filter (US, CA, IN, UK, Global)	Included
Fetch URL	Reads any public URL - web pages, PDFs, Office docs, images	Included
Search Documents	Re-queries your uploaded matter documents with new search terms	Included
Search Legal Corpus	Searches millions of US, Canadian, and Indian court cases with court and year filters	Included
Resolve Citation	Resolves citation strings (Bluebook, SCC, AIR, INSC) to full case details	Included
Generate Case Brief	Creates an IRAC case brief from a corpus judgment, optionally saves as DOCX	Paid
Court Records (eCourts)	Live search of 240M+ Indian eCourts records - case details, orders, cause lists	Paid (opt-in)

Additionally, any MCP server tools and Skills you have enabled are also available to the agent.

Read Any URL

The fetch_url tool lets the AI read content from any public URL. It handles:

Format	How It’s Processed
Web pages (HTML)	Main article content extracted, navigation and ads stripped
PDFs	Full text extraction with layout awareness for tables and columns
Office docs (DOCX, XLSX, PPTX)	Full text extraction
Images (JPEG, PNG, TIFF, BMP, WebP)	OCR runs automatically
Legacy formats (DOC, XLS, PPT)	Converted and extracted
Plain text, JSON, XML, CSV	Read directly

URL fetching is rate-limited to 20 requests per hour. Responses are capped at 5 MB and truncated to 15,000 characters before being sent to the AI.

Tool Configuration

Core tools (web search, fetch URL, search documents, corpus search, resolve citation) are always available when Agent Mode is on. Paid tools have individual toggles:

• Court Records (eCourts) - Opt-in toggle with per-action pricing (search: $0.002, case details:$0.006, order summary: $0.03, cause list:$0.01)
• Generate Case Brief - $0.02 per brief (LLM-intensive)

Limits

Limit	Value
Max iterations per response	5
Max tool calls per iteration	3
URL fetch rate limit	20/hour per user
URL fetch timeout	10 seconds
URL response size cap	5 MB
URL text truncation	15,000 characters
Web search results per call	5

Example Prompts

Agent Mode shines with multi-step research questions:

“Compare the privacy frameworks in GDPR and CCPA, citing specific provisions from both”

“Read this judgment URL and prepare a case brief: https://example.com/judgment.pdf”

“Find recent Supreme Court cases on arbitration clauses in employment contracts and check if any have been overruled”

“Search for contract termination clauses in my uploaded agreement and compare them against market standard”

Related

• Skills - Curated legal playbooks the agent can activate
• MCP Client - Connect external tools for the agent to use
• Document Search - How document retrieval works
• Case Law Research - Legal corpus search details