Intent Registry

The registry is the shared list of job names IICP nodes and clients use to understand each other. It turns “I need chat” or “I need embeddings” into stable identifiers software can match.

Canonical protocol vocabulary with lifecycle, schema, fixture and released-implementation evidence. Registry status is not a live capacity or provider-availability claim.

These identifiers are the shared vocabulary used by AI DNS-style resolution and the IICP agent discovery protocol.

Registry v1.4.0

Updated 2026-08-08

6 active · 17 total

Why these names matter

Clients ask clearly

They request a known job type instead of guessing provider-specific names.

Nodes advertise clearly

They list the work they are prepared to perform for the mesh.

The mesh can match them

The directory can return candidates that speak the same task language.

Registry governance

Intent URNs are protocol vocabulary, so additions should be stable, versioned, and backed by implementation evidence. Active means the registry requirements are met; it does not mean that a provider is online. Experimental and reserved entries remain visible without being presented as mature interoperability contracts.

Check Nodes or Stats for time-bound network availability.

Download JSON →Propose an intent →

Intent URN	Name	Description	Kind	Status
urn:iicp:intent:llm:chat:v1Input schema Output schema Release evidence	LLM Chat Completion	Multi-turn chat completion using the messages array format (OpenAI-compatible).Input schema Output schema Release evidence	inference	active
urn:iicp:intent:llm:completion:v1Input schema Output schema Release evidence	LLM Text Completion	Single-turn text completion from a prompt string.Input schema Output schema Release evidence	inference	active
urn:iicp:intent:llm:embedding:v1Input schema Output schema Release evidence	Text Embedding	Generate dense vector embedding(s) from input text.Input schema Output schema Release evidence	representation	active
urn:iicp:intent:code:lint:v1Input schema Output schema	Code Lint and Analysis	Static analysis and lint of code. Returns issues with severity and location.Input schema Output schema	analysis	experimental
urn:iicp:intent:code:optimize:v1Input schema Output schema	Code Optimization	Suggest or apply optimizations to code.Input schema Output schema	transformation	experimental
urn:iicp:intent:research:summarize:v1Input schema Output schema	Research Summarization	Summarize a long document or set of documents.Input schema Output schema	transformation	experimental
urn:iicp:intent:education:tutor:v1Input schema Output schema	Educational Tutoring	Provide tutoring, explanation, or guided problem-solving on a topic.Input schema Output schema	interactive	experimental
urn:iicp:intent:translate:v1Input schema Output schema	Text Translation	Translate text from one natural language to another.Input schema Output schema	transformation	experimental
urn:iicp:intent:classify:v1Input schema Output schema	Text Classification (deprecated)	Classify text into one or more categories from a provided set. DEPRECATED: use urn:iicp:intent:llm:classify:v1 for model-assisted classification with an optional user-facing explanation. This alias is retained for Phase 1 backward compatibility only.Input schema Output schema	analysis	deprecated
urn:iicp:intent:embedding:multimodal:v1Input schema Output schema	Multimodal Embedding	Generate embeddings from mixed text and image inputs.Input schema Output schema	representation	experimental
urn:iicp:intent:mcp:tools/call:v1Input schema Output schema	MCP Tool Call	Execute an MCP tool call through IICP transport. Reserved for the formal MCP binding so MCP-compatible tools can be addressed as mesh capabilities.Input schema Output schema	tool	reserved
urn:iicp:intent:llm:summarize:v1Input schema Output schema	LLM Summarization	Summarize a document or long text into a condensed form. Part of the Cooperative Inference Profile (Phase 5).Input schema Output schema	transformation	reserved
urn:iicp:intent:llm:classify:v1Input schema Output schema	LLM Classification	Classify text using a language model with an optional user-facing explanation. Complements urn:iicp:intent:classify:v1, which is label-only. Part of the Cooperative Inference Profile (Phase 5).Input schema Output schema	analysis	reserved
urn:iicp:intent:llm:rerank:v1Input schema Output schema	LLM Reranking	Rerank a list of candidate passages or results by relevance to a query using an LLM or cross-encoder. Part of the Cooperative Inference Profile (Phase 5).Input schema Output schema	ranking	reserved
urn:iicp:intent:audio:transcribe:v1Input schema Output schema Release evidence	Audio Transcription (Speech-to-Text)	Transcribe speech audio to text using OpenAI-compatible transcription payloads and Whisper-class backends. Audio is carried as base64 in the request and posted by SDKs as a multipart upload to compatible backends.Input schema Output schema Release evidence	transformation	active
urn:iicp:intent:audio:speech:v1Input schema Output schema Release evidence	Text-to-Speech (Speech Synthesis)	Synthesize speech audio from text using OpenAI-compatible speech payloads. The backend returns binary audio, and SDKs carry it back as base64 in result.audio for transport through the mesh.Input schema Output schema Release evidence	generation	active
urn:iicp:intent:safety:moderate:v1Input schema Output schema Release evidence	Content Moderation	Classify text for policy-violating content using OpenAI-compatible moderation payloads or compatible safety classifiers. The request and response are plain JSON; the backend may supply its own fixed moderation model.Input schema Output schema Release evidence	analysis	active

Machine-readable registry available at /.well-known/iicp-intents.json.