Sign in
Docs What's available

What's available

What you can run today, from the engine and server to HeapDB Cloud, the integrations and agent memory, and what's coming next.

Here's what you can run today and what's coming next. HeapDB is an event-ranking engine you can embed in your program or run as a network service, with a hosted option in early access.

Available now

Open source, Apache-2.0.

  • The engine: heapdb 0.5 (heapdb-core, pip install heapdb): a Rust engine with a Python API (PyO3), prebuilt wheels for Linux, macOS and Windows, CPython 3.10+. It runs in your process with JSONL snapshots. Results are identical to the earlier pure-Python versions, checked by differential tests. Benchmarks
  • The Python SDK and HTTP: the HeapDB API in-process, and heapdb.connect() (heapdb 0.6.0 and later) for a HeapDB server over HTTP.
  • The server: heapdb-server (heapdb-core main, 0.6.0): the same engine as a network service, with a write-ahead log and snapshots, single-node on local disk or clustered on S3 and DynamoDB, and an edge container image built from main. See Self-hosting.
  • Agent memory: heapdb-memory 0.1 (heapdb-memory, pip install "heapdb-memory[mcp]"): a memory SDK, an MCP server (heapdb-memory-mcp: remember, recall, mark_used, mark_ignored, forget, list_memories) and LangGraph / LlamaIndex adapters. Measured on LoCoMo; see AI agent memory.
  • Scoring with lag kernels, dose and rules; top-k with segment pruning
  • Feedback learning with proportional credit, shared priors, joint causes
  • EXPLAIN on every answer; calibration report

Early access

  • HeapDB Cloud: hosted databases with a connection URL, API keys and a web console, on the same data plane as heapdb-server.

Available on request

  • Integrations: GitHub, PostHog and a generic alert webhook (Sentry, PagerDuty, Datadog via templates). Likely culprit PRs, pre-merge risk and product-analytics reports from live events. They work today: ask us for access.

Coming next

  • TypeScript client (@heapdb/client), on heapdb-core main and coming to npm.
  • Native connectors for Sentry, PagerDuty, Datadog, GitLab and Linear, and a community connectors repository.
  • A persistent storage engine (LSM-backed segments), then sharding with cross-shard top-k.

The order is on the roadmap. For how to read a ranking, see Design scope.