What is the Agent?
What is the Agent?
The Corto Agent is a single place to ask legal questions and get answers that draw on your matter documents and legal research at the same time. Instead of switching between features, you ask one question and receive one grounded, cited response.
The All tab is the foundation for where Matter AI is heading — one-tab, growing capabilities. It gives lawyers, paralegals, and other legal professionals a single destination to conduct their legal work.
AT A GLANCE
One workspace - agentic chat, visible reasoning, and connected tools in the All tab. No switching between features.
Grounded in your matter - answers draw on your matter documents and cite their sources.
Calls the right tools automatically - engages LawY legal research when your question needs them, with more integrations coming soon.
What the Agent Draws On
What the Agent Draws On
The Agent always works from your matter information, the documents and correspondence in the matter. On top of that, it calls on connected tools when your question calls for them, with no manual switching required:
Matter documents & correspondence (always): retrieves and ranks the most relevant files in your matter, so every answer is grounded in your actual records.
LawY legal research: when a question needs the law, the Agent automatically engages LawY and returns cited, authoritative sources, including Acts, sections, cases and relevant legal articles.
Prompts: pre-built prompt templates are available in the Prompts tab to help you get started quickly with common legal tasks.
Model selector: choose your preferred AI model from the toolbar — options include Claude and GPT models across Instant and Thinking modes, depending on the speed and reasoning depth your task requires.
Features toggle: click + in the toolbar to turn on Research or other available tools before submitting your question.
MORE TOOLS COMING SOON
The All tab is built to grow. Additional MCP-powered integrations will be added over time, so you gain new capabilities without learning a new interface - one tab, growing capabilities.
How to Use the Agent - Step by Step
How to Use the Agent - Step by Step
Step 1 — Open the All tab
Create a new matter in the Matter List by clicking Create new matter. You can import matters from storage or quickly create one via OneDrive or SharePoint. Open the matter you are working on from the Matter List. From the matter toolbar, select the tab relevant to your task — All, Matter, Prompts, or Research — each offering a different mode of working within the matter.
Step 2 — Filter Your Data Source
Click + New to start a new conversation. You can choose to Chat about this matter, Chat about a document, or Chat about a folder to narrow the AI's focus — or Select a prompt to start from a pre-built template. Previous searches and conversations are listed below for easy access.
Option | Best used when… |
Chat about this matter | You need a broad overview of the entire matter (default setting) |
Chat about a document | You know the specific document(s) to draw from |
Chat about a folder | The matter is well organised into named folders, and you want to draw on information from that folder/s |
Select a prompt | You want to start from a pre-built template for common legal tasks |
Step 3 — Turn on the tools you need and choose your model
Click + in the toolbar and toggle on the tools relevant to your question — Research for LawY legal research. The Agent only calls a tool when your query needs it, so you stay in control of what it uses.
Select your Model — ChatGPT Instant, ChatGPT Thinking, Claude Sonnet, or Claude Opus.
Step 4 — Type your question
Type your question or request in plain language. You do not need special commands or keywords.
For example:
• “Prepare a letter of advice to the client on the contract interpretation issue in this matter, with relevant case law and legislation.”
• “Summarise the matter and identify the main legal issues.”
Step 5 — Answer any clarifying questions
If your question is missing key details, the Agent asks up to three clarifying questions before it starts — one at a time. Common questions include:
Who should the document be addressed to?
What length or format do you need?
Each question offers suggested options to select from. There is always a “Something else” field where you can type your own answer if none of the options fit. You can also click Skip to move past a question, and the Agent will note the unanswered item as an assumption in the output.
ANSWER ONCE
Your answers are saved for the rest of your session. You will not be asked again for the same matter.
Step 6 — Watch the Agent work
The Agent thinks out loud, giving you full visibility into how it reaches an answer — step by step, in real time. Rather than waiting for a final response and wondering how it was reached, you can watch the process unfold:
• It analyses your question and confirms understanding before starting, so it always solves the right problem.
• It runs multiple targeted searches across your matter’s files at once, showing each query as it executes.
• It fetches and ranks relevant documents, showing how many were retrieved and surfacing the most important ones.
• When your question needs the law, it engages LawY and returns cited authorities.
It shows the clarifying questions and the responses.
• It writes the final answer only after retrieval is complete, so the response is grounded in your actual documents.
Each step — Analysing, Researching, Searching, Fetching, Writing — is labelled, timestamped, marked Complete as it finishes, and fully collapsible, so you stay in control of the interface.
TRANSPARENT BY DESIGN
The visible reasoning trail gives you a reviewable record of how each answer was produced — useful for confidence, compliance, and audit readiness. You always know whether an answer is grounded in your documents or drawn from general reasoning.
Step 7 — Review the response
The Agent returns a structured response drawing on your matter documents and any tools it called. Citations appear inline — click any link to go directly to the referenced document, matter file, or legal source.
When LawY research is used, a Resources section at the end lists all external sources: legislation, leading cases, and legal articles, with direct links to authoritative sources.
ALWAYS REVIEW BEFORE USE
Review all output carefully before use in court documents or client communications. The Agent is designed to support, not replace, your professional judgment.
Step 8 — Refine if needed
You can continue the conversation to refine the output without starting again. For example:
• “Shorten the factual background section.”
• “Change the tone to be more formal.”
Step 9 — Preview, download or save the document
When the output is ready, the Agent pre-selects the most appropriate format for your document type or asks a clarifying question if it needs to confirm. Supported formats:
Format | Best for |
Chat text | Quick inline answers, no downloadable file |
Word (.docx) | Letters, memos, advice, submissions |
Filed documents, formal opinions | |
Excel (.xlsx) | Balance Sheets, Fee schedules, chronologies, data tables |
HTML | Interactive outputs — filterable chronologies, live-recalculating trackers, expandable family trees |
PowerPoint | Client or internal briefings |
You can generate multiple documents at once and save any or all of them directly to your LEAP matter.
Step 10 — Verify the Output
Confirm the accuracy of the document/s before finalising. All references are clickable. The Resources section lists all legal sources, including legislation and case law. The References section lists all documents referenced.
Agent Best Practices
Agent Best Practices
Use the All tab as a single destination to conduct AI legal work.
Turn on the right tools. Toggle on the tool features required. The Agent calls a tool only when your query needs it.
Be specific in your prompt. State the purpose and format of the document you require. Greater context produces more accurate output.
Engage with the clarifying questions. They are designed to reduce rework. Complete answers produce a stronger first draft and responses are retained for the session.
Use the visible reasoning to check grounding. Follow the step-by-step trail to confirm whether an answer is grounded in your documents or drawn from general reasoning.
Refine using follow-up turns. Tone, length, and structure can be adjusted in the same conversation without starting again.
Verify every citation. All references are clickable. Confirm accuracy before filing or issuing to a client. The Resources section lists all legal sources, including legislation and case law. The References section lists all documents referenced.
HTML Outputs. If you want an interactive result, describe the interactivity directly (eg, filtering, live recalculation, expand/collapse).