Integrations
polign_db plugs into the frameworks you build agents with and the platforms you deploy on. Each one has its own page with the install, a working example, and the limits.
Frameworks
Small open-source Python packages. They talk to the same server as every other client and need nothing extra on the server side.
LangChain
A vector store with metadata filters, MMR, and keyword and hybrid search.
Use with LangChain →LangGraph
Resume an agent after a crash or upgrade from its Recall records, beside your checkpointer.
Use with LangGraph →Pydantic AI
Agents that remember their users across runs, with typed facts from Recall and corrections that replace the old value.
Use with Pydantic AI →LlamaIndex
A vector store with translated metadata filters and the text, hybrid, and MMR query modes.
Use with LlamaIndex →LiveKit Agents
Voice agents that remember the caller, with typed facts from Recall.
Use with LiveKit →Vapi
Phone assistants that remember callers and carry corrected facts into the next call.
Use with Vapi →Agent runtimes
Recall as the memory of an agent platform, installed as a plugin.
OpenClaw
Fills the memory slot. A correction replaces the old fact, the history stays, and the agent can read memory as of any earlier time.
Use with OpenClaw →Storage for Recall
Recall keeps its memory in polign_db by default. It can also use a database you already run.
PostgreSQL + pgvector
Every Recall capability, agent resume included, in a PostgreSQL database you already run. Agent writes are checked against their lease in the same transaction.
Use with PostgreSQL →MongoDB
Every Recall capability on Atlas or your own replica set, with exact vector search and no search index to build.
Use with MongoDB →Qdrant
Typed memory with the same correction rules and history, kept in your Qdrant. No agent leases or keyword search.
Use with Qdrant →Deployment platforms
The same release binaries, packaged for where you already run things. Your data stays in your own bucket on every one of them.
AWS Marketplace
One EC2 instance and an S3 bucket in your account, launched from a CloudFormation template. The software is free.
Launch on AWS →Azure virtual machines
One Linux VM and a Blob Storage container in your subscription, reached through the VM's managed identity.
Install on Azure →Phala Cloud
A confidential VM template with storage encryption always on and a token-protected endpoint.
Deploy on Phala Cloud →Kubernetes
A signed Helm chart for any cluster, with cloud identity for bucket access.
Run on Kubernetes →Agent Sandbox
Agents in disposable sandboxes keep their Recall memory outside the pod, with no secret inside it.
Use with Agent Sandbox →Clients and tools
Claude Code and MCP
Plugins that give Claude memory across sessions and search over your own collections.
Claude Code plugins →Anything that can call Python can use the Python SDK directly, and every other language can use the HTTP API or the gRPC API. If you want a first-class integration, tell us which one.