[ AGENTS YOU OWN, ACTIONS YOU CAN AUDIT ]
Custom AI agents with scoped tools into your practice management or EHR system, human confirmation gates, and an append-only audit log of every action. Built by the team behind four open-source MCP connectors.
30 min with a co-founder. Rough quote in 3 business days.
4
open-source connectors on npm
100%
of agent actions audit-logged
4-8 wk
to production
[ ANATOMY OF AN AGENT WE SHIP ]
Anyone can wire a model to an API. What makes an agent deployable in a law firm or a clinic is everything around the model.
01
Scoped tools into your systems: read a matter, list documents, query the calendar. You choose which tools exist. The agent cannot touch what it has no tool for.
02
Claude via the API with training disabled and Zero Data Retention, another commercial model, or a self-hosted model on your infrastructure. Swappable by design.
03
Writes go through human confirmation. Client-facing output waits for attorney or clinician sign-off. The agent drafts; a person decides.
04
Every tool call, every input, every approval, logged append-only. Defensible for ABA Opinion 512, HIPAA audits, and your own ethics review.
[ WHAT WE BUILD ]
Not demos. Working agents on the systems firms and clinics already run.
The agent reads one matter or chart through the API and produces first drafts: consult memos, status reports, clinical documents. Drafts land back in the system for professional review before anything is sent.
Multi-step processing of case records: medical and fact chronologies, deposition summaries, damages tables, each entry cited to its source document. How the chronology pipeline works.
Calendar availability across the team, intake calls turned into matter shells with parties and deadlines extracted, follow-up tasks created. Every write behind a confirmation gate.
[ READ THE CODE FIRST ]
Four open-source MCP connectors, MIT licensed, on npm and the official Anthropic MCP Registry. They are the tool layer of the agents we build, public for anyone to inspect.
@oktopeak/clio-mcp. Matters, contacts, documents, tasks, calendar, billing, audit-log export. Hub page.
@oktopeak/mycase-mcp. The first and only open-source MyCase connector. Hub page.
@oktopeak/filevine-mcp for PI litigation and @oktopeak/intakeq-mcp for behavioral and allied health practices. Same audit-logged architecture.
[ WHERE WE DEPLOY ]
Agent architecture is general. Compliance is not. We build where the audit requirements are hardest and the data is most sensitive.
On Clio and MyCase. Privilege-safe architecture, ABA Opinion 512 audit logging, attorney sign-off gates.
Legal AI integration →On your EHR and practice management system. BAA-covered, Zero Data Retention, human verification before the chart.
Healthcare AI integration →[ FAQ ]
A single agent workflow on a system you already run (Clio, MyCase, an EHR) starts at $1,700. Multi-step agent builds with compliance architecture run $17K-$53K fixed price depending on the number of workflows, systems integrated, and audit requirements.
Discovery is free: a call, a draft scope, and a rough quote before you commit. Post-launch retainers start at $2,420/month.
A chatbot answers questions. An AI agent does work: it has scoped tools into your real systems, performs multi-step tasks (read the records, extract the facts, draft the document, write it back), and takes actions under defined controls.
The agents we build have three properties a chat window does not: scoped tool access you configure, human confirmation gates on writes, and an append-only audit log of every action.
Whichever fits your confidentiality requirements. Most builds run on Claude via the API with training disabled and Zero Data Retention. We also build on other commercial models, and for organizations whose policies rule out third-party AI processors we deploy self-hosted models on your own infrastructure.
The agent architecture (tools, gates, audit log) is model-agnostic; the model is a swappable component.
Compliance is architectural, not a policy document. For law firms: append-only audit logging of every agent action aligned with ABA Formal Opinion 512, attorney sign-off before anything client-facing, matter-scoped data access. For healthcare: BAA coverage across the vendor chain, Zero Data Retention on the model, human verification before anything touches the chart. We have passed 3 HIPAA audits on systems we built.
Yes, before you ever talk to us. We open-sourced four MCP connectors that give Claude agent tools into practice management systems: Clio (26 tools), MyCase (18 tools), Filevine, and IntakeQ. All MIT licensed on npm and the official Anthropic MCP Registry under the @oktopeak scope. Read the code, check the audit logging, install one in five minutes.
A single agent workflow on an existing system ships in 1-2 weeks. Custom multi-step builds take 4-8 weeks to production, including the compliance architecture and team training. Discovery before that is free.
Rarely, and only when the problem looks like our specialty: an agent acting on regulated or high-liability data under audit requirements. If you need a general marketing or e-commerce agent, a generalist shop will serve you better and cost less. Our value is the compliance architecture, which only matters when the data is sensitive.
[ GET STARTED ]
A call with a co-founder, not a sales rep. Bring the workflow that eats the most hours; we will tell you honestly whether an agent can do it, what the controls need to be, and what it costs.
