Integrate Cognee with 2,000+ apps and services

After adding text data to a Cognee dataset, trigger cognify to build a knowledge graph based memory from it

Run Cognee enrichment tasks over an existing dataset graph (or over custom data with custom extraction/enrichment tasks)

Create a dataset by name (returns the existing one if the name is already taken)

List the documents/files stored in a dataset, including their UUIDs for Forget, Update and Delete Data

List the datasets you have access to

Pipeline status together with in-flight progress (files completed / total, current stage), one item per dataset

Pipeline status per dataset (pending, running, completed, failed). Use it to poll after Run in Background.

Permanently delete a dataset and all its associated data

Remove a specific data item from a dataset while keeping the dataset intact

Delete a dataset, a single data item, or everything you own. Optionally clear only the graph and vector memory while keeping raw files.

Bridge the session memory of one or more sessions into a dataset's knowledge graph, so later recalls find what was said. Idempotent per session: a run only processes entries added since the previous one.

Query memory with any Cognee search type, optionally combining knowledge-graph hits with session Q&A and trace entries

Ingest text or a binary file into a dataset and build the knowledge graph in one call (add + cognify)

Store a typed session memory entry: a question/answer turn, an agent trace step, or feedback on an earlier answer

Replace an existing data item with new text or a file; the old version is deleted and the new one is ingested into the graph

Run a search query in Cognee memory engine

Get one session with its Q&A and trace entries, usage and cost

List sessions with activity, status and cost, newest first

Apply a previously created proposal to the skill (writes the new procedure)

Delete one skill (graph node and embeddings) from a dataset

List the skills ingested into a dataset

Fetch a proposal with its before/after procedures, rationale and confidence

Fetch one skill (including its full procedure) by ID

Ingest inline SKILL.md markdown as a dataset-scoped skill (no file upload)

Record a low-scoring skill run and create a skill-improvement proposal (not applied)

Run an AGENTIC_COMPLETION search that loads the given skill to review a task