mem0 Release Notes

What's New

mem0 whoami — print the active agent's default_user_id (the AGENTRUSH leaderboard identifier). Reads from local config, no network call.

mem0 agent-rush <add | search> — new subcommand group that wraps the platform's /v1/agent-rush/ endpoints for the 7-day AGENTRUSH game. Project routing is implicit; no flags exposed.

PII safety prompt on first interactive mem0 agent-rush add. Acknowledgement persisted in ~/.mem0/config.json under agent_rush.acknowledged_at. Non-interactive (agent) invocations print the warning to stderr without blocking.

New config field: agent_rush.acknowledged_at (ISO timestamp).

Changed

Agent-rush HTTP requests send X-Mem0-Mode: agent-rush header for telemetry split.

Full diff: cli-v0.2.5...cli-v0.2.7

What's New

mem0 whoami — print the active agent's default_user_id (the AGENTRUSH leaderboard identifier). Reads from local config, no network call.

mem0 agent-rush <add | search> — new subcommand group that wraps the platform's /v1/agent-rush/ endpoints for the 7-day AGENTRUSH game. Project routing is implicit; no flags exposed.

PII safety prompt on first interactive mem0 agent-rush add. Acknowledgement persisted in ~/.mem0/config.json under agent_rush.acknowledged_at. Non-interactive (agent) invocations print the warning to stderr without blocking.

New config field: agent_rush.acknowledged_at (ISO timestamp).

Changed

Agent-rush HTTP requests send X-Mem0-Mode: agent-rush header for telemetry split.

Full diff: cli-node-v0.2.5...cli-node-v0.2.7

Mem0 Platform v3 can now interpret time-aware memories and queries so assistants retrieve the right information for questions about the past, upcoming plans, and current state.

Time-aware search intent

• Queries like last week, upcoming, right now, and as of March 2025 return contextually appropriate results automatically
• Enabled by default — No per-request toggle required for v3 writes or searches
• Anchored relative queries — reference_date anchors relative search phrases for tests, backfills, and reproducible demos
• Normal response shape — Temporal reasoning affects ranking while preserving existing client response patterns

See Temporal Reasoning for usage details.

Per-project search-time ranking bias that boosts recently-touched memories and gently dampens stale ones. Off by default; opt in per project via the decay field on the project endpoint, or via client.project.update(decay=True) in the SDKs (Python v2.0.2 / TypeScript v3.0.3).

• Soft bias, never a filter. The scaling factor stays in 0.3×–1.5×. Decay can reorder candidates but never zeros them out — anything that surfaced before decay can still surface after.
• Reinforcement loop. Every memory returned in a search has its access history updated, so frequently-used facts naturally float to the top over time.
• Public score still clamped to [0, 1]. Existing API contract preserved; no client-side changes needed.
• v3 search only, fully reversible. See Memory Decay docs.

Mem0 OpenClaw Plugin (v1.0.11)

New Features:

• Skills-mode auto-setup: enableSkillsConfig() now runs automatically after onboarding — enables triage, recall (with reranking + keyword search), and dream consolidation with tools.profile = "full" and disables the built-in session-memory hook to avoid conflicts
• Memory runtime capability: Plugin now exposes runtime.getMemorySearchManager() and resolveMemoryBackendConfig() on the registered memory capability, enabling OpenClaw gateway to query memory status and backend config directly
• Dimension-aware collections: OSS wizard detects embedder dimension changes and creates a new collection (mem0_<dims>d) automatically, with a warning about old memories being inaccessible under the new embedder
• Tool documentation in skills: Both memory-triage and memory-dream SKILL.md files now include full tool reference sections listing all available tools with parameters

Improvements:

• Auto-capture and auto-recall default to enabled: autoCapture and autoRecall now default to true (was false). Manifest descriptions updated accordingly. Ignored in skills mode
• memory_update over delete+add: Skills now prefer memory_update for in-place edits — atomic and preserves edit history. Consolidation pattern updated: update best memory, delete redundant ones
• Search threshold lowered: Default searchThreshold reduced from 0.5 to 0.1 for broader recall. Removed hardcoded 0.6 recall-specific override — all searches now use the configured threshold
• Embedder dimension propagation: Vector store config auto-resolves dimensions from embedder config when not explicitly set. Syncs dimension and embeddingModelDims fields for Qdrant/PGVector compatibility
• Config file write safety: writeFullConfig() now re-reads and deep-merges the plugins section before writing, preserving installs and slots written by the OpenClaw gateway
• Additional embedder models: Added mxbai-embed-large (1024), all-minilm (384), and snowflake-arctic-embed (1024) to known embedder dimensions

Security:

• Bumped protobufjs to >=7.5.5 via pnpm overrides (GHSA-xq3m-2v4x-88gg) (#5012)

Fixes:

• Moved bootstrapTelemetryFlag() and removed ensureInstallRecord() from module-level side effects — both now run inside register() to avoid crashes when loaded outside OpenClaw gateway
• Fixed OSS history DB path resolution: absolute paths no longer passed through resolvePath(), preventing double-prefix bugs
• Manifest providerAuthEnvVars replaced with spec-compliant setup.providers format using id + envVars

Dependencies:

• Bumped mem0ai from 3.0.1 to 3.0.2
• Bumped pluginApi and minGatewayVersion compat to >=2026.4.24

Mem0 Python SDK (v2.0.1)

Bug Fixes:

Client: Map user_id, agent_id, run_id entity params to filters in GET /memories (#4960)

Memory: Honor prompt param in vector store extraction pipeline (#4914)

Memory: Add missing text_lemmatized field in AsyncMemory._create_memory (#4886)

Memory: Merge same-key operator dicts in AND metadata filters (#4853)

LLMs: Narrow _is_reasoning_model check to not match gpt-5.x variants (#4746)

Vector Stores: Add ca_certs config option for Elasticsearch vector store (#3993)

Vector Stores: Add agent_id and run_id to Elasticsearch/OpenSearch default mappings (#4906)

Embeddings: Set FastEmbed embedding_dims from model metadata at init (#4711)

Security:

Bump vulnerable dependencies to patched versions (#4835)

Mem0 OpenClaw Plugin (v1.0.10)

Security

Telemetry distinct_id now uses SHA-256 instead of MD5 — prevents rainbow-table reversal of API key hashes

User email is now SHA-256 hashed before sending as distinct_id — no PII in telemetry payloads

Declared PostHog telemetry endpoint (us.i.posthog.com) in providerEndpoints

Fixes

Fixed version-pinned install records preventing plugin updates. ensureInstallRecord() now detects semver-pinned specs (e.g. @mem0/[email protected]) and rewrites them to @latest or clawhub: prefix so openclaw plugins update resolves to the newest release

Fixed searchThreshold default inconsistency: standardized to 0.3 across docs, README, and manifest

PLUGIN_VERSION now injected at build time via tsup define from package.json — no more hardcoded version strings

Manifest Compliance

Removed non-spec fields: requiredEnvVars, dataLocations, privacy, setup (with externalEndpoints, providers, requiresRuntime, postInstallHint)

Replaced setup.externalEndpoints with spec-compliant providerEndpoints using endpointClass + hosts format

Env var declarations now rely solely on providerAuthEnvVars (already spec-compliant)

Docs

Fixed openclaw plugins update command: uses plugin ID (openclaw-mem0), not npm package name (@mem0/openclaw-mem0)

Added update section to README

Removed redundant "Key Features" and "Conclusion" sections from integration docs

Ground-up rewrite of the memory pipeline with 20+ point benchmark improvements:

• LoCoMo: 71.4 → 91.6 (+20) — multi-turn conversation recall
• LongMemEval: 67.8 → 93.4 (+26) — long-term memory across sessions
• BEAM (1M tokens): 64.1 — production-scale memory evaluation
• Agent memories are first-class — Previous algorithm: 46% on assistant recall. New: 100%
• Temporal reasoning works — “Where did I live before SF?” Previous: 51%. New: 93%
• ~3-4x fewer tokens — Under 7K tokens per retrieval vs 25K+ for full-context approaches
• ADD-only extraction — Memories accumulate; nothing is overwritten or deleted
• Hybrid retrieval — Semantic + BM25 keyword + entity boost, scored in parallel
• Entity linking — Entities extracted, embedded, and linked across memories

Breaking changes: Graph memory removed from OSS, search() defaults changed, deprecated params removed. See migration guide.

AI coding agents in Claude Code, Cursor, and Codex can now access Mem0 knowledge directly in their workflow — no doc searching required. Three interconnected skills launched:

mem0 Core Skill

Complete Python and TypeScript SDK reference, REST API patterns, and integration guides for LangChain, CrewAI, Autogen, and more

mem0-cli Skill

Terminal command reference, configuration walkthroughs, and CI/CD recipes

mem0-vercel-ai-sdk Skill

Vercel AI SDK provider API, memory-augmented generation patterns, and multi-provider setup

A full-featured command-line interface for Mem0, available in both Python and Node.js:

• Install: pip install mem0-cli or npm install -g @mem0/cli
• Full command suite — add, search, list, get, update, delete, import, config, init, status, entity, event
• Interactive setup — mem0 init with email verification or direct API key entry
• Works everywhere — Platform (Mem0 Cloud) and self-hosted OSS modes
• Scriptable — --json flag for CI/CD pipelines and automation
• Dual SDK — Same commands, same experience across Python and Node.js