Skip to content
Agents tracked: 501 Releases (7d): 777 Agent status: no reported issuesSince last limit reset: Claude 10d · Codex 8d For AI agents: Markdown Updated Nov 21, 2026

Memori changelog: what's new each month

Every stable Memori release summarised by month: the highlights, new features, improvements, fixes and anything you need to act on. 9 months covered; the current month updates daily.

May 2026

4 releases: 3.3.3 → 3.3.6

May brought documentation improvements, enhanced SDK examples, and backend refinements across Memori's TypeScript and Python support. The month focused on clarifying setup processes and strengthening multi-platform build capabilities.

Highlights

  • Merged Python and TypeScript BYODB documentation for clearer setup guidance.
  • Added fastuuid package and improved MongoDB ID handling for better data consistency.
  • Enhanced CI workflows to support multi-platform Rust builds and Android.
  • Added warning when allowConversationAccess is not configured to help users catch configuration issues.
  • Improved documentation verified against Python and TypeScript SDK source code.

New

  • fastuuid package added for enhanced MongoDB ID handling.
  • Claude Code Plugin added to Memori.
  • Pull request template added for improved contribution guidelines.
  • Agent trace and execution documentation added under core concepts.
  • Project-scoped MCP attribution guidance in docs.

Improved

  • Python SDK documentation sections now open by default in README.
  • CI workflows enhanced for better multi-platform Rust builds and Android support.
  • BYODB and Cloud documentation improved and verified against SDK source.
  • AI model examples updated in merged BYODB documentation.
  • Native Rust build path fixed for TypeScript.
  • Removed irrelevant providers from LLM Provider installation section.
  • Cargo arguments added for Intel Mac builds.

Fixed

  • Fixed requestheaders return type widening issue.
  • Fixed native Rust build path for TypeScript.
  • Fixed MySQL datetime type handling.
  • Removed unneeded console log from file.

April 2026

6 releases: 3.2.5 → 3.3.2

April brought quality-of-life improvements, expanded database support, and infrastructure enhancements to Memori. The month saw multiple releases focused on stability, configuration options, and platform compatibility.

Highlights

  • Added ability to delete entity memories with new deleteentitymemories method.
  • Expanded database support to include MongoDB with driver handshake metadata and TiDB.
  • Introduced Rust core with option to opt out for improved performance and flexibility.
  • Fixed tool-call corruption by sanitizing recalled history.
  • Enhanced platform support including macOS 15 Intel and improved Android compatibility.

New

  • deleteentitymemories method for removing entity-specific memories.
  • BYODB (Bring Your Own Database) configuration support.
  • MongoDB driver handshake metadata (DriverInfo) integration.
  • TiDB database driver support.
  • Question template for GitHub issues to improve user support.
  • Rust core implementation with opt-out capability.
  • Lazy Rust core initialization for flexible deployment.

Improved

  • LLM quality-of-life updates.
  • Logging configuration for sentence transformers.
  • Type annotations and docstrings for API documentation.
  • Embedding configurations to suppress progress output.
  • CI workflows for better test coverage management.
  • Package management and publish workflow processes.
  • Android support for Rust-backed Python extension.
  • Documentation including Memori Cloud and MCP overview.

Fixed

  • Suppressed transformers warnings during model loading.
  • Tool-call corruption by sanitizing recalled history.
  • Typo in banner (changed extraction to execution).
  • aiohttp dependency updated to version 3.13.2 for compatibility.

March 2026

3 releases: 3.2.2 → 3.2.4

March brought expanded documentation across integrations and benchmarks, along with new cloud recall capabilities and summary features.

Highlights

  • Added LoCoMo benchmark for performance evaluation.
  • New support for cloud recall summaries and bring-your-own-database summaries.
  • Expanded integration documentation for LangChain, Notion, and Slack.
  • Improved MCP setup guidance with dedicated README section.

New

  • LoCoMo benchmark added.
  • Cloud recall summaries support.
  • BYODB (bring-your-own-database) summaries.
  • DeepSeek provider documentation.
  • LangChain integration documentation.
  • Notion integration documentation.
  • Slack integration documentation.
  • Dedicated MCP section in README.

Improved

  • MCP setup guidance expanded and clarified.
  • Memori MCP documentation expanded.
  • OpenClaw documentation expanded.
  • LoCoMo benchmark documentation and paper references updated.
  • Fact creation now saves in human-readable format.

Fixed

  • Cloud recall update bug.
  • Gemini wrapper bug.
  • Broken links in documentation.
  • README syntax issues.
  • Minor documentation errors.

February 2026

2 releases: 3.2.0 → 3.2.1

February brought improvements to MongoDB compatibility, better error handling for LLM providers, and expanded Python version support. The month focused on stability fixes and API refinements.

Highlights

  • MongoDB ObjectId values are now properly supported in search operations.
  • Improved error handling for LangChain registration and unsupported LLM providers.
  • Added support for Python 3.13 and 3.14.
  • Fixed MongoDB migration issues with duplicate index keys.
  • Enhanced API subdomain handling and conversation message processing.

New

  • UnsupportedLLMProviderError for clearer error messaging when using unsupported language model providers.
  • Support for the new google.genai SDK in LangChain ChatGoogleGenerativeAI adapter.
  • Hosted metrics capability.

Improved

  • API subdomain handling now more robust.
  • Conversation message processing enhanced.
  • LangChain integration error handling made more informative.
  • Better error messages for common integration issues.

Fixed

  • MongoDB ObjectId conversion error in search functionality.
  • LangChain register error with improved error handling.
  • MongoDB migration IndexOptionsConflict caused by duplicate index keys.
  • Compatibility with new google.genai SDK in LangChain adapter.

January 2026

4 releases: 3.1.3 → 3.1.6

January brought expanded AI provider support and improved embedding capabilities to Memori. The SDK gained support for additional vector databases and LLM clients while refining its core search and context handling.

Highlights

  • Added support for Azure OpenAI and OpenAI Responses clients
  • Introduced support for OceanBase and SeekDB vector databases
  • Enhanced embedding configuration and error handling
  • Added context management and connection handling improvements
  • Improved debug logging and SDK documentation

New

  • Azure OpenAI client support
  • OpenAI Responses client integration
  • OceanBase and SeekDB vector database support
  • Context management and connection handling for Memori
  • Debug logging support

Improved

  • Embedding configuration and error handling
  • Google GenAI response handling
  • Embeddings and search utilities refactored for better performance
  • SDK documentation enhanced
  • Recall API improvements

Fixed

  • Google GenAI response handling

December 2025

6 releases: 3.0.4 → 3.1.2

December brought version 3.1.0 as a major update with improved memory augmentation capabilities and expanded provider support. The month focused on stability, documentation, and handling edge cases in asynchronous operations.

Highlights

  • Added graceful wait functionality for asynchronous memory augmentation to handle interrupts and termination cleanly.
  • Expanded LLM provider support including Agno, Nebius AI Studio, and enhanced Google GenAI integration.
  • Introduced new exception classes for better error handling in Memori API calls.
  • Added comprehensive troubleshooting guide and quickstart page to improve onboarding.
  • Refactored augmentation payload system with hashed entity and process ID tracking.

New

  • Agno provider support
  • Nebius AI Studio support
  • LLM instrument registration functionality
  • augmentation.wait() method for handling asynchronous memory augmentation
  • New exception classes for Memori API error handling
  • Issue templates for bug reports and feature requests
  • Troubleshooting guide documentation
  • Security policy

Improved

  • Test Mode now routes Advanced Augmentation tests to staging environment
  • Refactored augmentation payload to include hashed entity and process ID information
  • Enhanced Google GenAI driver capabilities
  • Automatic memory injection for Google GenAI provider
  • Documentation updates with simplified registration examples
  • Added Digital Ocean gradient deployment example

Fixed

  • Automatic memory injection for Google GenAI provider
  • Pymongo compatibility issue
  • Quota testing bug
  • Unsupported error handling

October 2025

1 release: 2.1.1

October brought improvements to MongoDB connection handling and reliability, particularly for Atlas deployments using DNS seedlist URIs.

Highlights

  • Fixed DNS resolution warnings when connecting to MongoDB Atlas with mongodb+srv:// connection strings
  • Enhanced SRV URI parsing to properly handle DNS seedlist discovery
  • Improved error handling for server topology inspection
  • Better type safety with MyPy type checking fixes
  • Refined conditional import patterns for optional dependencies

Improved

  • Connection string parsing logic for DNS seedlist discovery
  • Error handling for server topology inspection
  • SRV URI detection to avoid unnecessary DNS resolution attempts
  • Conditional import patterns for optional MongoDB dependencies
  • Hostname parsing in MongoDB connector and manager modules

Fixed

  • DNS resolution warnings on MongoDB Atlas connections
  • Hostname parsing logic in mongodbconnector.py and mongodbmanager.py
  • MyPy type checking errors for conditional MongoDB imports
  • Error handling for server descriptions without address attributes

September 2025

3 releases: 2.0.0 → 2.1.0

September brought Memori to version 2.0 with a complete architectural redesign, simplifying the memory system and separating short-term personalization from long-term memory. The month concluded with MongoDB support and improved debugging tools.

Action needed
  • consciousingest no longer controls long-term memory ingestion; it now only handles short-term personalization injection into system prompts.

Highlights

  • Long-term memory now works automatically without requiring special configuration.
  • Short-term personalization data now injects directly into system prompts instead of being stored as long-term memory.
  • MongoDB support added alongside existing SQL backends with automatic detection.
  • Performance improved through fewer API calls and optimized search logic.
  • Enhanced debugging with detailed logs for memory searches and error reporting.

New

  • MongoDB backend support with automatic detection of SQL or MongoDB database type.
  • Simplified memory architecture reducing complexity for users and developers.
  • Debug logging for memory searches, category filtering, and retrieval functions.
  • Stack traces and contextual error details for troubleshooting.

Improved

  • Long-term memory ingestion now happens automatically without manual configuration.
  • Short-term personalization context injected at system prompt level for better separation of concerns.
  • Database search logic improved for more reliable memory retrieval.
  • Error handling and reporting across different database backends.
  • Fewer API calls required for memory operations.

Fixed

  • Memory search resilience across different backend types.
  • Error transparency with improved stack trace reporting.
  • Database retrieval logic for more consistent results.

August 2025

2 releases: 1.0.0 → 1.0.1

Memori launched its first production release with comprehensive memory management for any LLM, featuring automatic context injection and multiple memory types. The month focused on establishing a solid foundation with universal LLM compatibility and structured memory processing.

Highlights

  • Works with any LLM library including LiteLLM, OpenAI, and Anthropic
  • Automatic context injection brings relevant memories into conversations without manual setup
  • Multiple memory types support short-term, long-term, rules, and entity relationships
  • Full-text search with semantic ranking helps find memories effectively
  • Pydantic-based validation ensures structured and reliable memory processing

New

  • Universal LLM integration compatible with any LLM library
  • Automatic context injection for relevant memories
  • Multiple memory type support: short-term, long-term, rules, and entity relationships
  • Advanced search with full-text and semantic ranking capabilities
  • Pydantic-based intelligence for structured memory processing
  • Configuration management with automatic settings loading

Improved

  • Modular design with separated concerns and clear component boundaries
  • SQL query centralization in dedicated modules for better maintainability

Summaries are written automatically from the official release notes (full changelog ↗); check the original notes before relying on a detail. Memori: pricing, features and alternatives · All changelogs