Check-in
AgentReads the insurance card from a photo and checks eligibility in real time.
Your teamConfirms coverage with the patient and collects the copay.
The AI harness for healthcare.
Clinicians and agents work side by side in Syntropic. Agents take on the charting, coding, and follow-up around each visit, and your team stays in control of every step.
- 01The clinician asks.
Dr. Lee suggests a skill in the encounter channel.
- 02The agent does the work.
It updates the chart and reports exactly what changed.
- 03The clinician decides.
Every section keeps its versions, and only the clinician signs off.
A real Syntropic encounter, shown with synthetic patient data.
One harness across the whole practice
Clubhouseteam + agent chat
Encounterscharting + scribe
Front deskcheck-in + coverage
Billingcoding + claims
Atlasclinic knowledge
Auditevery access, logged
Side by side
Agents do the work around care. Your team makes the calls.
Every workflow in Syntropic is shared. Agents prepare, check, and draft. People approve, decide, and sign.
Exam room
AgentDrafts each section of the note from the visit and applies your templates as skills.
Your teamThe clinician asks for changes, edits anything, and signs. Signing is never delegated.
Billing
AgentCodes the signed visit, explains each code against the note, and sends the claim.
Your teamBillers review the reasoning and work the exceptions.
Between visits
AgentSummarizes each visit on the Lounge cards and works in the same channels as your staff.
Your teamWhoever is free takes the next step, and anyone can see what the agent did.
Changes to the record wait for approval. Signing, orders, and closing encounters stay with clinicians. Every step lands in one audit trail.
Agents
The model is the easy part. The harness is everything around it.
What an agent can touch, what it knows, and when it has to ask. Your team sets all three, so agents help with real work without ever going around the people responsible for it.
You decide what every agent can do.
Each capability is read-only or side-effecting, scoped to a level of PHI, and gated by approval.
- Approval cards show the exact change before anything is written.
- Hard limits keep signing, orders, and closing encounters with clinicians.
- Untrusted evidence. Tool results inform an agent but never instruct it.
Grounded in your own protocols.
Ask Atlas answers from your clinic's knowledge base and cites every article it used.
See Atlas in a demo
Visit to claim
Side by side, from check-in to a sent claim.
Agents handle the routine steps. People handle the judgment calls, with every step the agent took in view.
Coverage confirmed before they sit down.
The agent reads the insurance card from a photo and checks eligibility in real time. Your front desk confirms the result and collects the copay.
The agent drafts the note. The clinician signs it.
Syntropic records the visit, drafts each section, and suggests codes. The clinician can change anything, and nothing is final until they sign.
Claims that file themselves.
When AI coding of a signed visit is complete, Syntropic sends the claim. Your billing team sees the reasoning and works the exceptions.
- Explained codes that point back to the note.
- A clear timeline from signature to submission.
Clubhouse
Your team and your agents, in the same channels.
The Lounge offers each next step to whoever is free, with an AI summary of the visit on every card. Agents post in the channels your staff already use, so anyone can see what they did and step in.
Tour Clubhouse
Bring your own agent
Every tool, over MCP and the command line.
Connect any MCP client or script to the same tools Syntropic's agents use, with the same permissions and the same audit trail.
syntropic-api schedule todayJSON + next steps
Tools across the practice
Scheduling, eligibility, charting, coding, claims, Clubhouse, Atlas, surveys, logs, and audit.
Scoped by role
OAuth scopes and per-position tool grants decide which tools each client can call.
Your choice of model
Anthropic, Amazon Bedrock, OpenAI, Gemini, and Vertex AI, limited to the providers you have approved for patient data.
Agents that learn your clinic
Teach workflows as skills and let agents remember preferences between conversations.
Questions about your data
Authorized, read-only SQL reporting for administrators, with every query logged.
Built for automation
Every response is JSON with suggested next actions, so scripts and agents always know what they can do next.
Security
Every call on the record.
People, agents, and MCP clients land in one audit trail with purpose of use and PHI level. Ask the Audit Assistant who looked at a chart, and why.
One audit trail for humans and agents.
See who acted, on whose behalf, through which tool, and how much patient information it touched.
Built for PHI from the first line
- Dedicated storage and encryption key for each organization ISOLATION
- Every tool declares the level of PHI it touches PHI TIERS
- Purpose of use recorded on every call AUDIT
- OAuth scopes and per-position tool grants ACCESS
Guardrails agents can't argue with
- Tool results are treated as evidence, never instructions INJECTION
- Side-effecting actions wait for a person to approve APPROVAL
- No signing, orders, releases, or closing encounters HARD LIMITS
- Access anomalies surfaced for review MONITORING
How we build
Agents draft. Clinicians decide. Anything that changes the record waits for a person, and the clinical sign-off is never delegated.
Every action is attributable. When an agent acts, you can see who asked, which tool it used, and what it read.
Least privilege by default. Agents and integrations get only the tools and the patient information their job requires.
Questions
What do you mean by an AI harness?
The model is only one part of an AI system. The harness is the rest: the tools an agent can call, the patient and practice context it can read, the permissions and approvals that bound it, and the audit trail that records it. Syntropic provides that harness for healthcare work.
Who stays in control when an agent works?
Your team. Administrators decide which tools each agent can use. Read-only capabilities run right away. Side-effecting capabilities, such as changing a chart section, wait for approval. Signing notes, placing clinical orders, releasing messages, marking results reviewed, and closing encounters are never available to agents.
Which AI models does Syntropic use?
Agents run on models such as Claude through Amazon Bedrock by default. Administrators can choose among supported providers, limited to those approved for use with patient data.
Can we connect our own agents and scripts?
Yes. Syntropic exposes its tools over MCP and a JSON command-line API. External clients are bound by OAuth scopes and per-position tool grants, and every call is audited.
How is patient information protected?
Each organization gets dedicated storage and its own encryption key. Every tool declares the level of PHI it touches, and every access records who, what, and why. We are happy to walk through our security practices in detail.
Are the screenshots on this site real?
The encounter at the top of this page is a real Syntropic screen. Before capturing it, we replaced every patient, staff, and clinic detail with synthetic data. The other screens are faithful recreations of the product, also filled with synthetic data. No real patient information appears anywhere on this site.
Give your agents a harness.
See your team and your agents working side by side, in your own workflows.