Mediforce is our open-source platform for agentic workflows in regulated pharma. Define a process end to end, decide which steps a person runs, a script runs, or an agent runs, and get an audit trail on every run, out of the box.
Request a demoRunning agents safely inside a pharma org is harder than running them anywhere else. Access control, team permissions, auditability, and compliance have to come first, and most platforms either skip that layer or bolt it on after the demo wins the deal. So you spend months building infrastructure scaffolding before an agent touches a real workflow, or you ship something fast that breaks the moment QA asks how a decision was made.
In Mediforce you define your process as a workflow, and every step is assigned to a person, a script, or an AI agent. Steps stay visible, trackable, and auditable. Autonomy is configurable per workflow: an agent can act on its own, or pause for human approval, and that is your call to make step by step, not a setting baked into the product.
The agents themselves are specialized components that each do one well-defined task: assess the quality of a data package, summarize a document, flag an anomaly. Because each agent does one thing, the same agent gets reused across different workflows. That reuse is where the value compounds.
On top of the platform, Mediforce ships pre-built workflow modules, including medical-monitoring triage, protocol-to-TLF, and a CRO data landing zone, with more in development. Start from a working module configured to your data and your process, or build your own with our team. Either way you skip the blank-slate phase that every project usually starts with.
Mediforce is model-agnostic by design. Connect Claude, OpenAI, Gemini, or another provider through your own API key. No lock-in to a single vendor's model, and no waiting for the platform to support the one you want.
Every step, every decision, every action is tracked and reviewable, built to meet the standards regulated work is held to.
Different roles see different agents and different access levels. Agents act independently or wait for approval, configured per workflow.
Built by a team that has shipped work for 8 of the world's 10 largest pharma companies. Mediforce reflects how pharma actually runs.
Connect Claude, OpenAI, Gemini, or another provider with your own API key. Your choice, not ours.
Mediforce stands up inside your environment with your auth, your team structure, and your permission model, so the regulated plumbing is in place before any agent runs.
Choose a process to automate, from the use-case library (medical-monitoring triage, protocol-to-TLF, CRO data landing zone) or a new one scoped with our team.
Set each agent's scope, access, and review gates per role. Decide where a human approves, where the agent acts alone, and where it escalates.
Move from one agent-assisted step to broader automation as the workflow earns trust. Never overnight, always with the audit trail intact.
Pharmaverse is an open-source R consortium used by pharmaceutical companies.
Hundreds of R packages in the ecosystem, with no systematic way to assess quality at scale.
A Mediforce agent evaluates each package's health and produces a structured assessment. A human council reviews that assessment and approves it or overrides it. The agent does the heavy lifting; the people keep final judgment.
A live, public example of how Mediforce runs in a regulated environment, the same human-in-the-loop pattern that a clinical or statistical workflow needs, running in the open where anyone can see how it works.
A workflow detects incoming CRO data the moment it lands, runs structural and content checks against the study spec, returns flagged issues to the CRO, and notifies the internal team only when something needs human judgment. The same headcount handles more studies, with a full audit trail per file.
Use-case description, not a named client engagement.Mediforce is built to support validation, not to sit outside it. Every run produces the audit record a regulated process needs: inputs, outputs, the model version, the tools called, and the person who reviewed and approved each step. Human review is part of the workflow itself, and model versions are tracked to the granularity 21 CFR Part 11 calls for. Your team still owns validation. Mediforce gives them an environment built to make it pass.
Our Validated AI approachAgentic workflows for regulated pharma: a person, a script, or an agent on each step, with an audit trail on every run.