Cognee Repo Skill

Use this skill when the user asks about Cognee, AI memory, persistent agent memory, graph RAG, Cognee SDK/CLI/API/MCP usage, Cognee configuration, custom graph extraction, session memory, or self-hosted Cognee services.

Cognee is a Python package and service stack for ingesting data, building a graph/vector-backed memory layer, and recalling knowledge for agents.

First checks

For a safe installed-package check:

python scripts/check_install.py --help
python scripts/check_install.py --json

Minimal package install:

python -m pip install "cognee"
python - <<'PY'
import cognee
print(cognee.__version__)
print(cognee.SearchType.GRAPH_COMPLETION)
PY

Do not assume provider credentials or optional databases are configured. Most real ingestion/search workflows need an LLM and embedding provider; backend-specific workflows need matching extras and services.

Route map

User intent	Read
Store, build, recall, improve, or delete memory with SDK calls	sub-skills/core-memory/SKILL.md
Choose SearchType, tune recall/search, handle empty results or invalid query knobs	sub-skills/search-retrieval/SKILL.md
Configure LLMs, embeddings, database/storage backends, optional extras, env vars, and paths	sub-skills/configuration-backends/SKILL.md
Define custom graph schemas, custom tasks/pipelines, ontology workflows, memify, migration, export, or visualization	sub-skills/advanced-graphs-pipelines/SKILL.md
Use session memory, feedback entries, agent memory decorators, agent identities, or multi-agent isolation	sub-skills/agent-session-memory/SKILL.md
Operate Cognee through CLI, FastAPI, MCP, Docker Compose, UI, cloud/local service connection, or push/sync	sub-skills/api-cli-services/SKILL.md

Shared references

• references/package-overview.md — package identity, public surface, and sub-skill boundaries.
• references/troubleshooting.md — cross-cutting install/import/credential/backend issues.
• references/repo-provenance.md — source snapshot and refresh baseline.
• references/repo-routing-metadata.json — structured router placement used by DisCo repo-skills import tooling.

Operating rules

• Prefer remember/recall for memory-shaped user requests; prefer add/cognify/search for explicit pipeline control.
• Before diagnosing workflow failures, separate three causes: package install, provider credentials, and database/storage backend state.
• Never tell a future agent to open original Cognee docs, examples, notebooks, tests, or scripts. Use the bundled references and scripts in this skill tree.
• Keep real API keys, cloud tokens, local filesystem paths, and service URLs in the user’s runtime configuration, not in copied guidance.
• Treat Docker, MCP, UI, and API servers as long-running services; run only --help, version, or check scripts unless the user asks to start them.
• If a current checkout has changed from the provenance snapshot, refresh this skill before relying on implementation-specific details.