Our take
Activeregular commits and releases- Merges more pull requests than 81% of the projects we track
EverOS is an actively maintained, Apache-2.0 Python memory runtime with a distinctive Markdown-first design, and its optional providers mean teams should match the setup to the retrieval and ingestion features they need.
Good fit if
- You want memory files that users can inspect, edit, and version, rather than state held only in a managed service.
- You need a shared memory layer across agents or apps and can run a Python 3.12+ service.
- You want to start with keyword retrieval and add embeddings, reranking, or multimodal support as needed.
Look elsewhere if
- Your deployment cannot meet the Python 3.12+ requirement or you specifically need a packaged Docker or Helm deployment, which the project does not list as a requirement or deployment option.
- You need hybrid search, reflection, or the Knowledge Wiki without configuring the additional embedding or reranking providers these capabilities require.
- You want office-document ingestion without installing LibreOffice, which EverOS uses to convert office files to PDF.
All health signals
| Last commit | 2026-10-01 (5 days ago) |
|---|---|
| Commits, last 90 days | 81 |
| Releases, last 12 months | 17 (latest v1.4.1, 2026-09-24) |
| Contributors | 8 (top contributor: 36% of commits) |
| Issues closed, last 90 days | 11+ (typically closed in 11 days) |
| Pull requests merged, last 90 days | 67 (typically merged in 0 days) |
| Project age | 11 months |
Checked on 2026-10-07 with the GitHub API.
Overview
EverOS is a local-first memory runtime for AI agents and the people who build with them. It addresses the problem of memory being trapped in an app, service, or database by keeping Markdown files as the readable source of truth and building local indexes for search.
The project separates user memory, such as episodes and profiles, from agent memory, such as cases and skills. Its Python library and API can serve multiple tools and workflows, with optional model providers enabling richer retrieval and memory processing.
Key Features
- Stores conversations, files, and agent trajectories as editable Markdown, with local SQLite and LanceDB indexes.
- Supports keyword search in its basic setup, with embedding-based and hybrid retrieval available when configured.
- Scopes retrieval across user, agent, app, project, and session identifiers.
- Watches Markdown changes and synchronizes indexes, so files can be edited directly.
- Provides a Knowledge Wiki with source-backed Markdown pages, taxonomy, and topic search when the required capabilities are configured.
- Offers offline reflection to consolidate episode clusters and refine profiles and skills between sessions.
- Integrates with agent platforms and workflow tools, including OpenClaw, Hermes, DeepSeek Harness, Dify, and Raven.
- Can ingest images, PDFs, audio, and office documents through an optional multimodal setup.
Use Cases
- Coding-agent teams can retain project context and decisions across sessions instead of relying on a single assistant's chat history.
- Agent developers can provide user-owned memory shared across several apps, tools, or workflows.
- Personal assistant builders can store preferences and past interactions locally, then retrieve them in later conversations.
- Workflow and integration authors can add explicit memory search and storage to agent platforms such as Dify or OpenClaw.
- Teams handling mixed content can add images, PDFs, audio, or office documents to memory when they configure the optional parser and multimodal provider.
What you need
Detected in the repository
- Python >=3.12 (from pyproject.toml)
- A test suite and automated checks on GitHub Actions
License in plain words
Apache-2.0permissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the license, the NOTICE file and a note of your changes
- Share your changes: no
- Includes an explicit patent grant from the contributors:
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Install the package and run the standalone demo, which does not require an API key:
pip install everos
everos demo
For server setup, optional providers, and API examples, see the README and documentation.
Alternatives
- OpenViking: OpenViking organizes agent memories, knowledge, and skills in a virtual filesystem, while EverOS stores conversations and context in editable Markdown.
- memoripy: Memoripy uses versioned, sourced records with explanations for recall; EverOS keeps agent conversations and context in Markdown with local indexes.
- Athena-Public: Athena carries coding context between AI sessions through a Markdown routine, while EverOS provides a Python memory runtime for agents across apps and sessions.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| EverOS | Python | Apache-2.0 | 13.4k | Active |
| OpenViking | Python | AGPL-3.0 | 39.2k | Active |
| memoripy | Python | Apache-2.0 | 694 | Active |
| Athena-Public | Python | MIT | 592 | Active |
Considerations
The measured project health is active: commits and releases are regular, and recent issue handling and pull-request merges indicate ongoing maintenance. Work is spread across multiple contributors rather than being concentrated in one person, but the contributor pool is still relatively small, so teams should assess whether its pace and support model suit their needs.
The basic memory flow needs Python 3.12 or newer and an LLM provider for the documented server setup. Embeddings and reranking are optional, but several features depend on them; multimodal ingestion needs its own configuration, and office files additionally require LibreOffice. The repository describes a local SQLite and LanceDB stack, not a requirement for a separate database service. Its Apache-2.0 license permits broad use subject to the license terms.
Found this useful?
Share it with someone who would like EverOS.