# hippo-memory

> Biologically-inspired memory for AI agents. Decay, retrieval strengthening, consolidation. Zero dependencies.

- **URL**: https://www.freshcrate.ai/projects/hippo-memory
- **Author**: kitfunso
- **Category**: RAG & Memory
- **Latest version**: `v1.22.0` (2026-06-03)
- **License**: MIT
- **Source**: https://github.com/kitfunso/hippo-memory
- **Language**: TypeScript
- **GitHub**: 583 stars, 25 forks
- **Registry**: github
- **Tags**: `typescript`

## Description

Biologically-inspired memory for AI agents. Decay, retrieval strengthening, consolidation. Zero dependencies.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v1.22.0` | 2026-06-03 | High | ### Fixed - **In-force E2 objects no longer drop out of the knowledge graph when their mirror memory is forgotten or consolidation-pruned.** Graph entity/relation provenance is now anchored to the authoritative E2 object (decision / policy / customer-note / project-brief) instead of the decaying memory mirror. Previously a decision or policy whose mirror memory decayed away (or was forgotten) silently vanished from `hippo graph extract` and graph recall, even while the object itself was still a |
| `v1.14.0` | 2026-05-28 | High | ## 1.14.0 (2026-05-28): J2 availability-bias detector (Track J)  ### Added  - **`availabilityHint` on `RecallResult` (Track J, J2).** A soft warning that   fires when a recall's returned top-K is dominated by recent entries (over 70   percent created in the last 24h) while substantially older relevant matches in   the same candidate pool were passed over. This is the availability / recency   heuristic (Tversky-Kahneman): what is most recent gets over-weighted relative   to what is most relevant. |
| `v1.10.1` | 2026-05-22 | High | Patch release closing the last open item in the v0.37 server-hardening cluster, deferred from v1.10.0.  ## Shipped  - **`stop()` no longer unlinks a newer server's pidfile.** `serve()`'s `stop()` and the `cli.ts` stale-pidfile self-heal both removed `server.pid` unconditionally. If a second server had started on the same hippoRoot and rewritten the pidfile, the first server's shutdown deleted the second one's pidfile and orphaned it. The new `removePidfileIfOwned` unlinks the pidfile only on a ` |
| `v1.8.1` | 2026-05-09 | High | Pre-commitment retraction patch. The v1.8.0 prereg's "Pre-committed v1.9 direction" — *"v1.9 will run the dlPFC goal-stack mechanism on the LongMemEval R@5 corpus as a cross-validation"* — is **RETRACTED publicly**. Outside-voice review on two iterations of the v1.9 plan (v1 and v2) found six structural barriers that preclude the mechanism from firing on the LongMemEval corpus + canonical harness as shipped, without substantial re-architecture. Per `CLAUDE.md` "Root Cause Over Patches" + the v1 |
| `v1.6.4` | 2026-05-05 | High | Two deferred items from the v1.6.2 senior review queue. The full review chain caught **5 bugs across 3 rounds** that any single reviewer would have missed:  ## Round 1 — plan-stage \`/codex\` + \`/review\` (caught 2 P0s pre-coding)  - Cross-tenant probe leak: my draft of \`drillDown\` had an unscoped \`readEntry\` to distinguish \`wrong_tenant\` from \`not_found\`. Would have leaked cross-tenant existence. - Validator ordering bug: \`validateIdSegment\` ran AFTER \`matchPath\` split — \`%2F\` de |
| `v0.36.0` | 2026-04-29 | High | ## A1 Server Mode - hippo serve, MCP-over-HTTP, thin-client CLI  ### Added - **`hippo serve` daemon** on http://127.0.0.1:6789 (configurable via --port or HIPPO_PORT). Exposes /v1/memories, /v1/auth/keys, /v1/audit, MCP-over-HTTP at /mcp, and /health. - **CLI thin-client.** When the daemon is running, CLI invocations auto-detect via .hippo/server.pid and route through HTTP. Stale pidfile self-heals on first ECONNREFUSED. - **MCP-over-HTTP/SSE transport** alongside the existing stdio path. POST / |
| `v0.31.0` | 2026-04-22 | High | **Route corrections to the right context.** Memories can now carry a scope tag, so a correction said during one skill surfaces strongly when that skill runs again, and stays quiet elsewhere.  ### Added  - **`hippo remember --scope <name>`** tags memories with a context scope. - **`scopeBoost` in search scoring** — 1.5x when the active scope matches the memory's scope, 0.5x when it mismatches, 1.0x for unscoped. Same multiplier pattern as the existing `decisionBoost` / `pathBoost` / `outcomeBoost |
| `v0.24.2` | 2026-04-16 | High | ## What's new\n- Added a machine-level daily runner that sweeps registered Hippo workspaces and runs daily learn + sleep without requiring one OS task per project.\n- Updated the OpenClaw plugin to queue detached session-end autosleep so consolidation no longer blocks shutdown.\n- Updated the docs to separate query-time retrieval from session-end and daily refresh behavior across integrations.\n\n## Verification\n- npm run build\n- npm test\n- npm publish |
| `v0.3.1` | 2026-04-15 | High | ### Added - `hippo init` auto-detects agent frameworks (Claude Code, Codex, Cursor, OpenClaw) and installs hooks automatically. Use `--no-hooks` to skip. - `hippo learn --git --repos <paths>` scans multiple repos in one pass (comma-separated paths). - Codex integration guide (`integrations/codex.md`).  ### Changed - README rewritten with auto-hook install docs, multi-repo learn section, and updated comparison table. - OpenClaw integration guide updated with auto-install instructions and m |
| `v0.3.0` | 2026-04-15 | High | ### Added - Cross-tool import: `hippo import --chatgpt`, `--claude`, `--cursor`, `--markdown`, `--file`. - Conversation capture: `hippo capture --stdin` / `--file` (pattern-based, no LLM). - Confidence tiers: `--verified`, `--observed`, `--inferred`. Auto-stale after 30 days. - Observation framing: `hippo context --framing observe\|suggest\|assert`. - All import commands support `--dry-run`, `--global`, `--tag`. - Duplicate detection on import. |

## Citation

- HTML: https://www.freshcrate.ai/projects/hippo-memory
- Markdown: https://www.freshcrate.ai/projects/hippo-memory.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/hippo-memory/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._