EverOS
✓ Text source-checked Nov 22, 2026Open-source memory OS for AI agents with Markdown files, sub-500ms retrieval, and self-evolving skills.
Source-checked against the vendor's site · rebuilt from the vendor's own pagesProfile updated Link checked How we rate
In short
EverOS is an open-source memory OS for AI agents that uses Markdown files for sub-500ms retrieval. It features self-evolving skills and supports both cloud and self-hosted deployment.
EverOS syncs local SQLite and LanceDB indexes from the Markdown files to enable fast retrieval.
EverOS auto-extracts and tags each memory as Profile, Episodic, or Skill without manual curation. Repeated successful agent tasks self-promote into reusable Skills shared across the agent team. EverOS ingests PDFs, images, docs, excels, slides, and URLs in a single call. EverOS uses an mRAG system to deliver precise memory retrieval at inference speed without loading the entire context window.
EverOS is compatible with Claude Code, OpenClaw, Hermes, Codex, MCP clients, and the OpenAI and Anthropic SDKs. Plugins are available for DeepSeek Harness and Dify, and EverOS is built into Raven.
Users can start on EverOS Cloud and move to self-hosted deployment at any time. The EverOS repository is available on GitHub under the Apache 2.0 license. The vendor says that all memories export as clean Markdown, ensuring data portability and version control.
According to EverOS: EverOS syncs local SQLite and LanceDB indexes from the Markdown files to enable fast retrieval. Its pages list this strength: Self-evolving skills that promote repeated wins into reusable Skills.
— The AI Tool Bible editorial team
Pros
- ✅ Self-evolving skills that promote repeated wins into reusable Skills.
- ✅ The vendor says that 94.74% accuracy on the LoCoMo benchmark.
- ✅ Sub-500ms p95 retrieval latency.
- ✅ Multimodal ingestion of PDFs, images, docs, excels, slides, and URLs.
- ✅ Open source with an Apache 2.0 license.
- ✅ Markdown-first data export for portability.
Facts
- Platforms
- linux, macos, windows, web, cli (not source-checked)
- API available
- Yes
- Free trial
- Yes (not source-checked)
- Open source
- Yes
- Earliest Wayback capture
- 2025-12-09 (earliest archived capture of the site, not a launch date)
Where this tool sits
Use cases
Frequently asked
- How much does EverOS cost?
- EverOS offers a freemium model. The Cloud Free plan is $0.00/month with 3 Memory Spaces. The Cloud Pro plan costs $24.99/month and includes 8 Memory Spaces. Users can also self-host the open-source repository under the Apache 2.0 license.
- Which AI tools and SDKs does EverOS integrate with?
- EverOS is compatible with Claude Code, OpenClaw, Hermes, Codex, MCP clients, and OpenAI and Anthropic SDKs. Plugins are available for DeepSeek Harness and Dify, and the system is built into Raven for seamless agent integration.
- Is EverOS difficult to set up or learn?
- EverOS auto-extracts and tags memories as Profile, Episodic, or Skill without manual curation. It ingests various file types in a single call. Users can start on the cloud and move to self-hosted deployment at any time, ensuring flexibility.
- Can I export my data from EverOS?
- Yes, all memories export as clean Markdown files. This ensures data portability and allows for version control. The system syncs local SQLite and LanceDB indexes from these Markdown files to enable fast retrieval.
- What file types can EverOS ingest?
- EverOS ingests PDFs, images, docs, excels, slides, and URLs in a single call. It uses an mRAG system to deliver precise memory retrieval at inference speed without loading the entire context window.