How AI agents, agent frameworks and MCP servers connect to case data, which tools and models work with Prima.Law, and 10 secure ways to run your law office with AI.
Connect Claude, Gemini, ChatGPT or local models through n8n, Zapier, MCP or your own code.
Only firm administrators generate API keys, alongside user roles and library access.
AES-256 encryption at rest, encryption in transit, and data centers in the United States.
Your prompts, case data and AI outputs are never used to train public or commercial models.
Two years ago, most law firms were still deciding whether to try artificial intelligence at all. That question has been answered. In the American Bar Association’s 2024 Legal Technology Survey, 30% of attorneys said their firms used AI-based tools, up from 11% the year before. At firms with 100 or more lawyers, the figure jumped from 16% to 46%. Clio’s 2025 Legal Trends Report found that 79% of legal professionals now use AI in some part of their work.
The next shift is already under way. AI is moving from chatbots you type questions into, to AI agents: software that takes a goal, looks up the information it needs, uses tools, and completes multi-step work, with a person reviewing the result. Three trends are driving it:
For a law firm, all of this raises one practical question: how does an AI agent safely reach your case data?
An AI agent is only as useful as the information and actions it can reach. An agent that triages new leads needs intake details and a way to create a task. An agent that drafts client updates needs to know when a case status changes. An agent that keeps the books current needs invoice and payment events.
Without an API, firms fill that gap by hand: staff copy case details into a chat window, paste the answer back, and hope nothing confidential ends up somewhere it shouldn’t. That is slow, error-prone and impossible to supervise at scale.
An external API (application programming interface) solves this the right way. It gives approved software a controlled doorway into your system, with defined events it can listen for and defined actions it can take. The firm decides who gets a key, which tools connect, and what each workflow may do. Client data stays in your case management system instead of being scattered across chat histories.
No two firms run the same way. One immigration practice captures leads from Facebook ads and a Spanish-language intake form; another books every consultation through Calendly. One reports to partners in Slack, another in Microsoft Teams. Some reconcile in QuickBooks Online; others manage hundreds of employer-sponsored cases with their own reporting. Off-the-shelf integrations cover the common cases. The firms that grow fastest need their own intake logic, their own prompts, their own tools and their own rules about who sees what. That is where an open API becomes part of how the business runs.
“AI agent” covers several layers. Knowing them helps a firm choose the right tool for each job.
| Layer | What it does | Examples |
|---|---|---|
| AI model | Reads, reasons and writes | Claude (Anthropic), Gemini (Google), GPT models in ChatGPT (OpenAI), open models run locally with Ollama |
| Agent framework | Code libraries for building agents that plan, call tools and hand off work | LangGraph, CrewAI, OpenAI Agents SDK, Claude Agent SDK, Google ADK, Microsoft Agent Framework |
| Workflow automation | Connects apps and runs steps when events happen, now with AI steps | n8n (cloud or self-hosted), Zapier |
| Tool protocol | A standard way for AI assistants to discover and call tools | Model Context Protocol (MCP) servers and clients |
| System of record | Where client and case data lives, with permissions | Prima.Law, through External API Access |
n8n is a workflow automation platform that firms can use in the cloud or host on their own servers. Its AI Agent node connects to chat models from Anthropic, OpenAI, Google and others, including local models. For MCP, n8n offers two building blocks:
Paired with the Prima.Law API, n8n lets a firm build custom agents that react to Prima.Law events, call the model of its choice, and write results back as tasks and notes for staff to review.
Zapier connects Prima.Law to thousands of apps without code, using the Prima.Law app in Zapier. Zapier MCP goes a step further: it exposes the Zapier actions you choose to MCP clients such as Claude and ChatGPT, so an assistant can take approved actions across your apps. Firms choose exactly which actions are available, and Zapier handles sign-in and app connections.
Firms with developers, or working with a technology partner, can build agents in code with frameworks like LangGraph, CrewAI, the OpenAI Agents SDK, the Claude Agent SDK or Google ADK, and connect them to Prima.Law through External API Access. This suits firms with their own data warehouse, internal portal or reporting needs.
The Prima.Law API is model-agnostic. Prima.Law does not lock your firm into a single AI vendor. Because agents reach Prima.Law through a standard API, n8n, Zapier or MCP, your firm can:
Inside Prima.Law itself, AI Case Review runs on Google Cloud’s Gemini Enterprise in an isolated enterprise environment, and your data is never used to train models.
Lawyers have duties most software users don’t. The ABA’s Formal Opinion 512 (July 2024), its first ethics opinion on generative AI, reminds lawyers that competence includes understanding the tools they use, that confidentiality applies before client information goes into any AI system, and that lawyers must supervise AI output. Attorneys feel this: in the ABA survey, 75% named accuracy and 47% named data privacy and security as top concerns about AI.
Before connecting an AI agent to client data, a firm should be able to answer these questions:
External API Access is a Prima.Law Enterprise feature, built for firms that want the most up-to-date features, deeper access, automation and AI agents to run their practice. It opens Prima.Law to the tools your firm chooses while keeping control with your administrators.
Each example keeps a person in the loop and gives the agent only the access it needs.
Yes. External API Access is available on the Prima.Law Enterprise plan. Administrators generate API keys in the Administrator Panel and connect Prima.Law to Zapier, n8n, AI agents and the firm’s own systems.
Yes. Prima.Law is model-agnostic. Firms connect their preferred model through n8n, Zapier, an agent framework, or an MCP server built with n8n’s MCP Server Trigger or Zapier MCP.
The Model Context Protocol (MCP) is an open standard that lets AI assistants discover and use tools. With n8n’s MCP Server Trigger or Zapier MCP, a firm can expose approved Prima.Law actions, such as creating a task or note, to assistants like Claude and ChatGPT.
It can be, with the right controls: admin-issued keys, least-privilege workflows, business-tier AI models that don’t train on your data, human review and a record of what agents did. Prima.Law encrypts data with AES-256, stores it in the U.S., and never uses your data to train AI models. Once data is sent to another provider, that provider’s terms for your plan apply, so verify them first.
Not necessarily. Many workflows can be built in Zapier without code, and n8n offers visual workflows with AI agent steps. Firms with developers can go further with agent frameworks and direct API access.
Technology will keep changing faster than any single software release can keep up with. An open, secure API lets your firm adopt new AI models, agent frameworks and tools as they arrive, on your terms, without giving up control of confidential client information. If your firm needs that flexibility, the Prima.Law Enterprise plan with External API Access is built for you.
Data in Prima.Law. Client and case data stored in Prima.Law is protected as described in the Prima.Law Privacy Policy, including encryption in transit and at rest and storage in the United States. Prima.Law’s built-in AI features, such as AI Case Review, process data under the enterprise terms described in that policy, and that data is never used to train AI models.
Data sent outside Prima.Law. When your firm uses the Prima.Law API, Zapier, n8n, an MCP server, an agent framework or any other tool to send Prima.Law data to a third-party AI model or service, that data leaves Prima.Law’s environment. How it is stored, retained, shared or used, including whether it may be used to train AI models, is governed by that provider’s terms, privacy policy and data-processing agreements, not by Prima.Law.
Terms depend on your license and subscription. AI providers often treat data differently depending on the product and plan: free or consumer accounts, paid individual plans, business or enterprise plans, and API access can each carry different data-use, retention and training terms, and settings within an account can change them. Your firm, and each user, is responsible for verifying what a provider will do with your data under your specific license, subscription and configuration before sending client information to it, and for re-checking when terms or plans change.
Your professional obligations. Each firm remains responsible for its own ethical and legal duties, including client confidentiality, informed consent where required, supervision of staff and technology, and the review of all AI-generated output before it is relied on or sent to a client, court or government agency. AI output can be inaccurate or incomplete.
No legal advice; third-party products. This article is general information about technology and is not legal advice. References to third-party products such as Claude, Gemini, ChatGPT, n8n, Zapier and the agent frameworks named above are for illustration only. Prima.Law does not control, and is not responsible for, third-party services, their features, availability or data practices, which may change at any time. Product names are trademarks of their respective owners.
Sources: American Bar Association, 2024 Legal Technology Survey Report (as reported by LawSites, March 2025); Clio, 2025 Legal Trends Report; ABA Standing Committee on Ethics and Professional Responsibility, Formal Opinion 512 (July 29, 2024); Linux Foundation, Model Context Protocol; n8n documentation (MCP Server Trigger, MCP Client Tool); Zapier MCP documentation; Prima.Law Privacy Policy.
Book a demo and tell us what you want to automate. We will show you how the Prima.Law Enterprise plan fits.