I built an AniList connector for Cognee so my AI agent remembers my anime watchlist
Description and use
The author built an AniList connector for Cognee so an AI agent can remember the user’s anime watchlist, scores, and deletions, enabling queries about preferences.
Problem
The author’s anime watchlist was only in their head or on AniList, which cannot answer questions, so they needed an AI agent with memory of watched shows, scores, and drops.
Outcome
The connector successfully syncs public AniList lists into Cognee, preserving edits and deletions, and enables the AI agent to answer queries about the user’s anime preferences.
Workflow
- 1. Identify a suitable data source (AniList) after MyAnimeList API failures.
- 2. Create a DLT source under packages/connector/anilist/ that pages through a user’s public list via AniList GraphQL.
- 3. Assign stable document IDs (e.g., anilist:21) for consistent node identity across syncs.
- 4. Implement edit‑survival logic so score changes propagate without duplication.
- 5. Implement delete‑forget logic so removed entries are dropped from memory.
- 6. Write and run tests (test_pipeline_sync_edit_forget) to verify edits and deletes.
- 7. Run a live smoke test against the author’s public AniList account (luffyimu).
- 8. Use the populated memory to ask the AI agent questions about ratings, preferences, and dropped shows.
Evidence
I built a data‑source connector that feeds a public AniList list straight into Cognee's memory layer.
Two behaviors I'm proud of, because they're the ones a reviewer actually checks: Edits survive re-syncs. Change a score on AniList, re-run, and the updated value flows through instead of duplicating the entry.
Deletes get forgotten. Remove something from your list and the next sync drops it from memory — forget-on-delete, the way a memory layer should behave.
The live smoke test ran against my own public AniList account (luffyimu). Output: TOTAL ROWS: 1 anilist:21 | ONE PIECE
- Record date: 2026-10-25T12:26:13.607Z