Memory for AI agents
- make informed decisions.

Recall memories are records to look back on

A memory, even for an agent, shouldn't be something that gets overwritten or lost. Memories are factual records of events and decisions. When agent needs the information to make a judgement call, memory should be at the center of it.

Recall memory is trusted, corrective and not made up

Have you ever had your agent hallucinate or just provide an answer when it doesn't know anything? Most of the memory setup today is simply a vector search or similarity search for whatever query that was put in. Recall looks for the exact facts were at any point in time. It provides enough context behind the ask to the agent. When agent has enough information about the situation, it can make an informed decision and generate predictable outcome for the user.

Compared with mem0 and Zep

Maya tells her assistant twelve things over nine months: three homes, three jobs, a sister who moves, and, in one sentence, a new diet and a shellfish allergy. Then she asks it to forget the allergy. The same model, Claude Opus 5.5, extracts and answers for all three memories, and every reply below is from a real run.

Use the buttons or the step tabs to go through it one step at a time. mem0 had stored the diet and the allergy as one memory, so forgetting the allergy deleted her diet too, and it answered from an older note that she is vegetarian. Zep deleted both of its diet facts. Recall withdrew only the allergy and answered pescatarian, with the vegetarian diet it replaced. Over three runs Recall answered 18 of 18, mem0 15 and Zep 11, and Recall used 3.2k tokens to save a message where mem0 used 12.7k and Zep 14.8k.

mem0 2.2.1 (open source), Zep's open-source engine graphiti-core 0.31.1, and Recall 0.20.2 with its built-in embedding model. Each answer was judged against a reference answer written before the runs. All three scenarios, with the numbers.

Try it in a minute

One pip install brings the client and the database. No model or embedding API key needed. The quick start walks through history, as-of reads, and connecting your agent.

pip install polign-recall

Runs on the store you already have

Recall is the memory. The vector store underneath is yours to choose, and the records are the same on every one. polign_db is the backend we built for it. A PostgreSQL or MongoDB you already run has every capability too, agent resume included, and a Qdrant works for the memory tools.

On polign_db your memory sits in your own S3, GCS, or Azure bucket. Searches straight from S3 measured about $3.20 per million, and nothing runs when agents are idle. See the method. What each backend supports, feature by feature, is in the architecture docs.

Resume after a crash

An agent can record its working state and turns to Recall as it works. When its pod dies or its image is upgraded, the next process picks up from those records instead of a snapshot. Read how it works, or use it with LangGraph.

How it is built

Every memory is an event in an append-only log, and every answer is a fold over that log at the instant you ask about. The registry holds the correction rules, the backend holds the records, and recall serve puts keys and namespaces in front of both. Read the architecture, with diagrams.

Become a design partner

Give us your agent and we will wire Polign Recall into it.

• You bring an agent that is in or near production, and a real memory problem
• We do the integration work, architecture help, a Recall deployment in your cloud on polign_db or the store you already run, and migration from your current memory store
• The commitment is one engineer on your side who owns it, and one 30-minute engineering call a week for the first 4 to 6 weeks
• Done means success criteria we write down together before day one, for one production-like agent writing memory into Polign
• Your infrastructure stays in your account and on your bill
• No charge for 60 days, with written success criteria agreed up front. Meet them and it converts to $12k for the next 12 months. See pricing

What ships today

• Go library with no dependencies outside the standard library
• Python client on PyPI as polign-recall
• MCP server for Claude Code, OpenClaw, and any MCP host, read-only or read-write
• One server for many agents, with keys bound to a namespace, and Kubernetes workload identity so a pod needs no key at all
• Claude Code plugin for memory across sessions
• LangGraph package for agents that resume after a crash, with a checkpointer and store in Recall
• OpenAI Agents SDK package for sessions whose history any process can read
• Pydantic AI package for agents that remember their users across runs
• LiveKit Agents package for voice agents that remember the caller
• Vapi adapter for caller memory through authenticated webhooks
• PostgreSQL, MongoDB and Qdrant backends for teams that already run one
• Durable tool calls: a resumed agent knows which calls a crash left unfinished, and never repeats one blindly
• Use cases: corrective memory, resume after a crash, context offload, shared memory across agents and an audit trail, each with a runnable example
• Comparisons: Mem0, Zep and Letta, LangGraph checkpointers, pod snapshotting and Temporal, the same eight rows each
• Starter predicates for coding and personal assistants; new ones are defined by first use, or keep a fixed list
• History, as-of reads and diff: how a fact changed, what was believed on any date, and what changed between two dates, from the CLI, MCP, or HTTP
• Audit bundles with an offline verifier
• Word-overlap search built in; model embeddings optional
• Plain text in, sentences out, optional: with a model configured, Recall works out the facts and decides how each is filed. Without one, your agent proposes facts and Recall validates them
• Explain shows why each answer is held: the text it came from, the rule that decides it, and its history
• Evidence links: a fact proposed from text keeps the exact words it came from
• Corrections across predicates: the registry pairs predicates that contradict each other (likes and dislikes, owns and gave away), so a later statement under one ends the same value under the other. An agent can name anything else a statement ends. Conflicts that no rule settles are shown, never guessed
• Forget is a withdrawal that keeps the history readable; delete the records in the store to erase them
• Runs where you run it: your laptop, your cluster, your bucket