Skip to main content
Agent Chat pulls context from multiple sources to deliver accurate, relevant results. Understanding what the agent can access helps you set up sessions that produce better output with less back-and-forth.

​
Context sources

When you start a conversation, the agent can draw from:
  • Your document: The full text of the contract open in Word, including tracked changes, comments, and revision history
  • Playbook: Your configured rules, fallback positions, and review criteria
  • Company profile: Your organization’s standard positions, preferences, and negotiation guidelines
  • Knowledge base: Your uploaded policies, templates, and reference materials
  • Uploaded documents: Files you attach to the conversation for reference (supports .docx, .pdf, .csv, and .xlsx)
  • Web search: External sources for case law, statutes, regulatory updates, and market standards
  • Filevine repository: Discovery documents, deal room files, and precedent from your connected Filevine projects
  • Conversation history: Everything discussed in the current session, including your edits and clarifying answers
The agent combines these sources automatically. For example, when you ask it to review a contract against your playbook, it reads the document, loads your playbook rules, applies your company profile preferences, and references any attached files without you needing to specify each one.

​
Setting up your session

Configure these settings before or during a conversation to shape what the agent has access to.

​
Playbook

Select a playbook to give the agent your review criteria, fallback positions, and standard language. The agent uses these rules when evaluating clauses and generating suggestions. If you don’t select a playbook, the agent still reviews based on general best practices and any context you provide in the conversation.

​
Company profile

Your company profile includes your organization’s standard positions, preferred terms, and negotiation guidelines. This context is applied automatically so you don’t need to repeat it in every prompt. To get the most out of this, keep your profile up to date with your current standards and preferences.

​
Review scope

Control what part of the document the agent reviews:
ScopeWhen to use
Entire documentFull contract review from start to finish
Tracked changes onlyReviewing counterparty redlines or recent edits
Specific clausesFocusing on particular sections you’ve selected
You can also highlight specific text in Word before prompting. The agent focuses on the highlighted passage rather than the entire document. Enable web search to let the agent look up case law, statutes, regulatory updates, and market standards during the conversation. This is useful for compliance reviews, benchmarking contract terms, and researching jurisdiction-specific requirements. Web search is off by default. Toggle it on in the session settings when your task requires external research.

​
Reference documents

Attach additional files to give the agent more context. Common uses:
  • Prior redlined versions of similar contracts
  • Your standard templates for comparison
  • Regulatory checklists or compliance requirements
  • Internal policies or guidelines
Supported formats include .docx, .pdf, .csv, and .xlsx. Files are available for the duration of your chat session.

​
Filevine documents

If your organization uses Filevine, connect your repository to search across thousands of discovery and deal room documents. The agent can pull relevant precedent, facts, and language into your drafting workflow. Select a Filevine project and specific documents from the session settings to narrow the search scope.

​
Providing context in your prompt

Beyond session settings, include relevant details directly in your message:
  • Your role: Vendor, customer, licensor, licensee
  • Document type: MSA, NDA, DPA, employment agreement
  • Jurisdiction: If specific laws or regulations apply
  • Risk tolerance: Light-touch review vs. aggressive negotiation
  • Audience: Who the output is for (internal team, counterparty, executives)
Example: 
Review this SaaS vendor agreement as the customer. We're in healthcare
under Delaware law. Focus on HIPAA compliance, indemnification, and
liability. We have moderate leverage in this deal.

​
How context affects results

The agent produces better output when it has the right context from the start:
  • Select your playbook and profile before the first message so the agent applies your standards from the beginning
  • Attach reference documents up front rather than mentioning them mid-conversation
  • Be specific about your role and jurisdiction in your first prompt so the agent doesn’t need to ask
  • Enable web search only when needed to keep the agent focused on your documents and playbook rather than general research