Uploaded August 2026 | Updated September 2026, 2 weeks ago
Every conversation about AI agents in clinical work ends up in the same place. Who approves the output, what happens when the agent gets something wrong, and whether QA will accept any of it.
Mediforce is our answer. It is an open-source platform for human-in-the-loop AI workflows in regulated clinical processes, now at v1.0.0 and Apache 2.0 on GitHub. A v1.0.0 means the platform is stable enough to build workflows on. It does not make anyone validated. Your validation process stays yours, and the evidence it asks for is already there when you go looking.
How it works
A workflow is a graph of steps, and every step runs in a Control Mode the platform enforces:
• No agent: a human, script, or automated action, with no AI involved
• Assist: the human does the work and AI reviews it afterward
• Cowork: agent and human work the step together, by chat or voice
• Human review: the agent completes the step and a person approves before the workflow moves on
• Autonomous agent: the agent runs the step and the workflow advances, reviewable later in the audit trail
Because it is set per step, the step that summarizes validation findings can run on its own, while the step that touches a submission waits for a named reviewer.
What else is inside
• A workflow assistant that builds with you. Describe the process and it lays the steps out on the canvas, and it shows you every change before the workflow saves.
• Bring your own key and set the model per step. Every run totals its own cost, so you can see which step is expensive and move it to a cheaper model from a dropdown.
• An audit trail that fills in as you work: who acted, in what role, what the agent read and produced, which model produced it, and the workflow version.
• Versioned, locked workflows and a dry run mode, so every run traces back to the exact definition that produced it.
Links
Website: mediforce.ai
Repository: github.com/Appsilon/mediforce
v1.0.0 release: github.com/Appsilon/mediforce/releases/tag/v1.0.0
Request a demo: appsilon.com/contact-us
Case studies
CRO data delivery: mediforce.ai/case-studies/data-delivery
Collecting submission documents: mediforce.ai/case-studies/collecting-documents
More from the Mediforce series
Mediforce | CDISC AI Innovation #1: youtu.be/ez3D9APC7aA
Mediforce | CDISC AI Innovation #3: youtu.be/_aKa__wX-iE
#Pharma #ClinicalTrials #OpenSource #AIagents #Biostatistics #CDISC #Appsilon
Every conversation about AI agents in clinical work ends up in the same place. Who approves the output, what happens when the agent gets something wrong, and whether QA will accept any of it.
Mediforce is our answer. It is an open-source platform for human-in-the-loop AI workflows in regulated clinical processes, now at v1.0.0 and Apache 2.0 on GitHub. A v1.0.0 means the platform is stable enough to build workflows on. It does not make anyone validated. Your validation process stays yours, and the evidence it asks for is already there when you go looking.
How it works
A workflow is a graph of steps, and every step runs in a Control Mode the platform enforces:
• No agent: a human, script, or automated action, with no AI involved
• Assist: the human does the work and AI reviews it afterward
• Cowork: agent and human work the step together, by chat or voice
• Human review: the agent completes the step and a person approves before the workflow moves on
• Autonomous agent: the agent runs the step and the workflow advances, reviewable later in the audit trail
Because it is set per step, the step that summarizes validation findings can run on its own, while the step that touches a submission waits for a named reviewer.
What else is inside
• A workflow assistant that builds with you. Describe the process and it lays the steps out on the canvas, and it shows you every change before the workflow saves.
• Bring your own key and set the model per step. Every run totals its own cost, so you can see which step is expensive and move it to a cheaper model from a dropdown.
• An audit trail that fills in as you work: who acted, in what role, what the agent read and produced, which model produced it, and the workflow version.
• Versioned, locked workflows and a dry run mode, so every run traces back to the exact definition that produced it.
Links
Website: mediforce.ai
Repository: github.com/Appsilon/mediforce
v1.0.0 release: github.com/Appsilon/mediforce/releases/tag/v1.0.0
Request a demo: appsilon.com/contact-us
Case studies
CRO data delivery: mediforce.ai/case-studies/data-delivery
Collecting submission documents: mediforce.ai/case-studies/collecting-documents
More from the Mediforce series
Mediforce | CDISC AI Innovation #1: youtu.be/ez3D9APC7aA
Mediforce | CDISC AI Innovation #3: youtu.be/_aKa__wX-iE
#Pharma #ClinicalTrials #OpenSource #AIagents #Biostatistics #CDISC #Appsilon
