> ## Documentation Index > Fetch the complete documentation index at: https://developer.suki.ai/llms.txt > Use this file to discover all available pages before exploring further. # Integration Prompts > Copy-ready prompts that point your AI coding agent at the right Suki skill, environment, and completion check for each workflow, SDK, and API Each prompt on this page starts an AI coding agent on one Suki integration. It tells the agent which skill to load, which environment to build against, and how you check the result, so the agent spends less time guessing and more time on your code. Prompts work best after you [Install agent skills](https://developer.suki.ai/documentation/references/agent-skills.md) and connect the [Documentation MCP](https://developer.suki.ai/documentation/references/mcp.md). The prompt names the skill to load, and the MCP lets the agent look up exact fields while it works. ## What each prompt includes | Part | What it tells the agent | | :- | :- | | Skill | The `SKILL.md` URL for the product, so the agent loads the documented workflow before it writes code | | Credentials | Where your Partner ID and Partner Token come from, and to keep tokens out of version control | | Environment | Build against staging unless you ask for production | | Done when | The observable result that shows the integration works | | Rules | Check every endpoint, field, and option against the docs, mark anything missing as **BLOCKED** instead of inventing it, and report untested cases as **Not run** | > [!NOTE] > > Prompts do not include credentials. Get your Partner ID, register your JWKS, and confirm template IDs through [Partner onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) before you run generated code. ## How to use these prompts 1. Choose the workflow, SDK, or API you are integrating. 2. Copy the prompt, or open it in Cursor from the prompt actions. 3. Paste it into your agent chat, and add details about your stack (language, framework, EHR) on the line after the prompt. 4. Review anything the agent marks as **BLOCKED** or **Not run** before you ship. ## Choose a workflow Use a workflow prompt when you know the job (Ambient notes, Form filling, Dictation) but have not chosen between an SDK and the Partner APIs. The agent loads the workflow skill, which routes to the right SDK or API skill. **Ambient** Load the Ambient skill, build against staging, and stop at a retrieved note. Build ambient clinical documentation with Suki for Partners. Fetch the Ambient skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-ambient/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-ambient/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: build against staging unless I ask for production. Direct APIs use [https://sdp.suki-stage.com](https://sdp.suki-stage.com) and wss\://sdp.suki-stage.com. SDKs take a staging environment option. Done when: an ambient session reaches `completed` in staging and the app retrieves the generated note content. Rules: check every endpoint, SDK method, field, and message type against the skill and the linked docs. Do not invent fields. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Form Filling** Load the Form filling skill, build against staging, and stop at retrieved form data. Build Form filling with Suki for Partners. Fetch the Form filling skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-form-filling/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-form-filling/SKILL.md) Credentials: use my partnerId from Suki partner onboarding, a partnerToken JWT from my identity provider, and the medical form template IDs Suki provides. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: build against staging unless I ask for production. Direct APIs use [https://sdp.suki-stage.com](https://sdp.suki-stage.com) and wss\://sdp.suki-stage.com. SDKs take a staging environment option. Done when: a Form filling session reaches `completed` in staging and the app retrieves the structured form data. Rules: check every endpoint, SDK method, field, and message type against the skill and the linked docs. Do not invent form fields or template IDs. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Dictation** Load the Dictation skill, build against staging, and stop at returned transcript text. Build clinical dictation with Suki for Partners. Fetch the Dictation skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-dictation/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-dictation/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: build against staging unless I ask for production. Direct APIs use [https://sdp.suki-stage.com](https://sdp.suki-stage.com) and wss\://sdp.suki-stage.com. SDKs take a staging environment option. Done when: the app receives transcript text from a Dictation session in staging. Rules: Dictation streams on /ws/transcribe, not /ws/stream. Check every endpoint, SDK method, field, and message type against the skill and the linked docs. Do not invent fields. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Patient Summary** Load the Patient Summary skill, build against staging, and stop at a retrieved summary. Build Patient Summary with Suki for Partners. Fetch the Patient Summary skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-patient-summary/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-patient-summary/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Use the same `sdp_suki_token` from Login for CKG FHIR upload and Patient Summary endpoints. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: build against staging ([https://sdp.suki-stage.com](https://sdp.suki-stage.com)) unless I ask for production. Done when: FHIR ingestion succeeds, summary generation reaches `COMPLETED` in staging, and the app retrieves the summary. Rules: check every endpoint, field, and status value against the skill and the linked docs. Do not invent FHIR resources or fields. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Audio Streaming** Load the audio streaming skill plus the product skill for the session you stream. Implement or debug Suki Partner audio streaming over WebSocket. Fetch the audio streaming skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-audio-streaming/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-audio-streaming/SKILL.md) Also load the product skill for the session you are streaming (Ambient, Form filling, or Dictation): [https://developer.suki.ai/.well-known/agent-skills/index.json](https://developer.suki.ai/.well-known/agent-skills/index.json) Environment: stream to wss\://sdp.suki-stage.com unless I ask for production. Ambient and Form filling use /ws/stream. Dictation uses /ws/transcribe with different message shapes. Done when: audio streams to the right endpoint in staging, the stream closes cleanly, and the session reaches its completed status. Rules: check every message type and field against the wire format docs. Do not invent message types. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Webhooks** Load the webhooks skill and stop at a receiver that verifies signatures. Build a Suki partner webhook receiver. Fetch the webhooks skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-webhooks/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-webhooks/SKILL.md) Configure my HTTPS callback URL during partner onboarding: [https://developer.suki.ai/documentation/webhook/configuration](https://developer.suki.ai/documentation/webhook/configuration.md) Credentials: keep the webhook secret on the server. Do not hardcode it or commit it to version control. Done when: the receiver verifies the HMAC signature on the exact raw request body before it parses JSON, and rejects requests that fail verification. Rules: check every header, event type, and payload field against the skill and the webhook docs. Do not invent event types. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) ## Choose your SDK Use an SDK prompt when you have already chosen the SDK. Each prompt names the option that points the SDK at staging. **Web SDK** Load the Web SDK skill, set the staging environment, and run an Ambient session end to end. Build with the Suki Headed Web SDK. Fetch the Web SDK skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-web-sdk/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-web-sdk/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: set environment: "staging" on SukiAuthManager unless I ask for production. The default is production. Done when: a clinician can run an Ambient session end to end in staging and the app handles the submitted note. Rules: check every component, prop, and option against the skill and the Web SDK API reference. Do not invent props or options. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Headless Web SDK** Load the Headless Web SDK skill, set the staging env, and control a session with hooks. Build custom Ambient UI with the Suki Headless Web SDK. Fetch the Headless Web SDK skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-headless-web-sdk/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-headless-web-sdk/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: create PlatformClient with env: "staging" unless I ask for production. Done when: the app starts, pauses, resumes, and submits an ambient session in staging with useAmbientSession. Rules: check every hook, type, and option against the skill and the Headless Web SDK API reference. Do not invent hooks or options. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Mobile SDK** Load the Mobile SDK skill, set the stage environment, and retrieve a note on device. Build Ambient documentation with the Suki Mobile SDK on iOS. Fetch the Mobile SDK skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-mobile-sdk/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-mobile-sdk/SKILL.md) Credentials: use my Partner credentials from Suki partner onboarding and a TokenProvider that returns a valid Partner Token. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: set SukiAmbientCoreManager.shared.environment = .stage unless I ask for production. Done when: the app records an ambient session on a device in staging and retrieves the generated note content. Rules: check every class, method, and property against the skill and the Mobile SDK API reference. Do not invent APIs. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Dictation SDK** Load the Dictation SDK skill, set the staging environment, and land text in a field. Embed Dictation with the Suki Dictation SDK. Fetch the Dictation SDK skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-dictation-sdk/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-dictation-sdk/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: set environment: "staging" on SukiAuthManager unless I ask for production. The default is production. Done when: dictated text from the hosted iframe lands in the target field in staging. Rules: check every class, prop, and option against the skill and the Dictation SDK API reference. Do not invent props or options. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Form Filling SDK** Load the Form filling SDK skill, set the staging environment, and receive filled values. Embed Form filling with the Suki Form filling SDK. Fetch the Form filling SDK skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-form-filling-sdk/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-form-filling-sdk/SKILL.md) Credentials: use my partnerId from Suki partner onboarding, a partnerToken JWT from my identity provider, and the medical form template IDs Suki provides. Do not hardcode tokens or commit them to version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: set environment: "staging" on SukiAuthManager unless I ask for production. Done when: a Form filling session completes in staging and the app receives the filled form values. Rules: check every class, prop, and option against the skill and the Form filling SDK API reference. Do not invent form fields, template IDs, or options. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) ## Choose your API Use an API prompt when you call the Partner APIs directly. Each prompt points the agent at the public [OpenAPI spec](https://developer.suki.ai/api-reference/openapi.json) so it can check request and response shapes. **Ambient API** Load the Ambient API skill, call staging, and stop at retrieved note content. Integrate the Suki Ambient Partner APIs (REST and /ws/stream). Fetch the Ambient API skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-ambient-api/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-ambient-api/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Keep sdp\_suki\_token on the backend. Never put it in client-side code, logs, or version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: use [https://sdp.suki-stage.com](https://sdp.suki-stage.com) and wss\://sdp.suki-stage.com unless I ask for production. Done when: the app logs in, creates an ambient session, streams audio over /ws/stream, ends the session, waits for `completed`, then retrieves the note content. Rules: check every endpoint, header, field, and message type against the skill and the OpenAPI spec ([https://developer.suki.ai/api-reference/openapi.json](https://developer.suki.ai/api-reference/openapi.json)). Do not invent fields. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Form Filling API** Load the Form filling API skill, call staging, and stop at retrieved form data. Integrate the Suki Form filling Partner APIs (REST and /ws/stream). Fetch the Form filling API skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-form-filling-api/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-form-filling-api/SKILL.md) Credentials: use my partnerId from Suki partner onboarding, a partnerToken JWT from my identity provider, and the medical form template IDs Suki provides. Keep sdp\_suki\_token on the backend. Never put it in client-side code, logs, or version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: use [https://sdp.suki-stage.com](https://sdp.suki-stage.com) and wss\://sdp.suki-stage.com unless I ask for production. Done when: the app logs in, creates a Form filling session with template IDs, streams audio over /ws/stream, ends the session, waits for `completed`, then retrieves the structured form data. Rules: check every endpoint, header, field, and message type against the skill and the OpenAPI spec ([https://developer.suki.ai/api-reference/openapi.json](https://developer.suki.ai/api-reference/openapi.json)). Do not invent form fields or template IDs. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Dictation API** Load the Dictation API skill, call staging, and stop at the final transcript. Integrate the Suki Dictation Partner APIs (REST and /ws/transcribe). Fetch the Dictation API skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-dictation-api/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-dictation-api/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Keep sdp\_suki\_token on the backend. Never put it in client-side code, logs, or version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: use [https://sdp.suki-stage.com](https://sdp.suki-stage.com) and wss\://sdp.suki-stage.com unless I ask for production. Done when: the app logs in, creates a transcription session, streams audio over /ws/transcribe, and reads `final_transcript` after the session reaches `completed`. Rules: Dictation uses /ws/transcribe, not /ws/stream. Check every endpoint, header, field, and message type against the skill and the OpenAPI spec ([https://developer.suki.ai/api-reference/openapi.json](https://developer.suki.ai/api-reference/openapi.json)). Do not invent fields. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) **Patient Summary API** Load the Patient Summary API skill, call staging, and stop at a retrieved summary. Integrate the Suki Patient Summary Partner APIs (CKG FHIR ingestion, generate, status, and retrieve). Fetch the Patient Summary API skill and follow it: [https://developer.suki.ai/.well-known/agent-skills/suki-patient-summary-api/SKILL.md](https://developer.suki.ai/.well-known/agent-skills/suki-patient-summary-api/SKILL.md) Credentials: use my partnerId from Suki partner onboarding and a partnerToken JWT from my identity provider. Use the same `sdp_suki_token` from Login for CKG FHIR upload and Patient Summary endpoints. Keep it on the backend. Never put it in client-side code, logs, or version control. [https://developer.suki.ai/documentation/get-started/partner-onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md) Environment: use [https://sdp.suki-stage.com](https://sdp.suki-stage.com) unless I ask for production. Done when: the app uploads a FHIR bundle, ingestion succeeds, generation reaches `COMPLETED`, and the app retrieves the summary. Rules: check every endpoint, header, field, and status value against the skill and the OpenAPI spec ([https://developer.suki.ai/api-reference/openapi.json](https://developer.suki.ai/api-reference/openapi.json)). Do not invent FHIR resources or fields. Mark anything you cannot find as BLOCKED, and report any test you did not run as Not run. Connect the documentation MCP for page search: [https://developer.suki.ai/documentation/references/mcp](https://developer.suki.ai/documentation/references/mcp.md) ## Next steps **[Agent skills](https://developer.suki.ai/documentation/references/agent-skills.md)** - Install every product skill with one command. **[Documentation MCP](https://developer.suki.ai/documentation/references/mcp.md)** - Live doc search in Cursor, VS Code, Claude, and Codex. **[FAQ for AI agents](https://developer.suki.ai/documentation/references/ai-agents-faq.md)** - Environments, the OpenAPI spec, and 401 errors. **[Partner onboarding](https://developer.suki.ai/documentation/get-started/partner-onboarding.md)** - Get your Partner ID and credentials. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.