sgx-labs/statelessagent

Overview

A local-first AI tool that gives agents persistent memory across sessions, packaged as a single binary with no API keys required. It solves the problem of AI forgetting context between interactions by maintaining state locally.

Best for

Best for
Developers seeking a simple, self-hosted memory solution for AI agents without cloud dependencies

Use cases

• Provide persistent context to AI chat assistants across multiple sessions
• Run offline AI agents without external network dependencies
• Deploy a lightweight, self-contained memory layer for custom tools

How to use

Install

curl -fsSL https://statelessagent.com/install.sh | bash

Tools exposed

• search_notes
• search_notes_filtered
• search_across_vaults
• get_note
• find_similar_notes
• get_session_context
• recent_activity
• save_note
• save_decision
• create_handoff
• reindex
• index_stats
• mem_consolidate
• mem_brief
• mem_health
• mem_forget
• mem_restore
• mem_list_suppressed
• save_kaizen

Tested with

Claude Code, Cursor, Windsurf, Codex CLI, Gemini CLI, Open WebUI, LobeChat

Example client config

{\n  "mcpServers": {\n    "same": {\n      "command": "npx",\n      "args": ["-y", "@sgx-labs/same", "mcp", "--vault", "/path/to/your/notes"]\n    }\n  }\n}

Notes

A local-first AI tool that gives agents persistent memory across sessions, packaged as a single binary with no API keys required. It solves the problem of AI forgetting context between interactions by maintaining state locally.

19 stars on GitHub. Last updated 2026-04-08.

Use cases

• Provide persistent context to AI chat assistants across multiple sessions
• Run offline AI agents without external network dependencies
• Deploy a lightweight, self-contained memory layer for custom tools

Pros

• Local-first approach ensures privacy and no ongoing API costs
• Single binary simplifies installation and deployment
• No API keys eliminates external service dependencies

Cons

• Very early stage (19 GitHub stars) with limited community support and polish
• Scalability and performance under heavy use are unproven
• May lack integration with popular AI frameworks or providers

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.