Mem0 vs Honcho

Mem0 reads each new message and pulls out candidate facts. Its April 2026 release made extraction ADD-only: "Memories accumulate; nothing is overwritten or deleted." The same release removed external graph databases from the open-source package in favor of built-in entity linking, and added BM25 keyword search beside the vector index (changelog). Adds are processed asynchronously: the call returns an event ID you poll until the memories are written (add memories).

When a user corrects something, Mem0 keeps both versions with timestamps and its retrieval layer decides which one to surface (Mem0's comparison guide). Several open issues track cases where the outdated value still wins, including #5867, where a maintainer called it "a design/feature gap", and #7535, which notes search ranking has no recency term.

Mem0 meters both writes and reads. The free plan includes 10,000 adds and 1,000 retrievals a month on one project. An agent that checks memory on every turn uses up the free reads at about 33 turns a day, across all its users (pricing).

Honcho stores each message as it was sent, labelled with the peer who said it. A peer is any participant: a person or an agent. A background process called the deriver reasons over new messages and writes conclusions about each peer, including things that follow from what was said but were never stated outright. When a user corrects a fact, the deriver writes a new conclusion with a timestamp, and the old one stays for now. Dreaming runs later, once enough new conclusions have built up, deletes conclusions a newer fact has replaced, and merges ones that say the same thing.

Your agent reads memory two ways. session.context() returns the representation, a summary, and recent messages inside a token budget, in about 200 milliseconds, and those reads are unlimited. peer.chat() takes a question and reasons over the representation to answer it, at one of five levels from minimal to max, priced per query from $0.001 to $0.50.

Writes are asynchronous. A conclusion lands after the deriver runs, in time for the next turn but not the same one. Each peer can also keep its own view of every other peer it has observed. A support agent's model of a customer and the customer's model of that agent are separate objects.

If your agent touches patient health data, Mem0's managed platform is SOC 2 Type 1, HIPAA, and GDPR compliant (Mem0). Honcho has SOC 2 Type 1 and doesn't offer HIPAA.

If your company's open-source policy excludes copyleft, Mem0's Apache-2.0 SDK clears review and Honcho's AGPL-3.0 code doesn't.