Description
Building things with an AI assistant, the constant pain is that context keeps getting lost — a day later, a different repo, a new chat, and you re-explain the project background and past decisions, while it sometimes proceeds on wrong assumptions. OpenContext gives the assistant persistent memory: a lightweight personal context / knowledge store that saves your background, decisions and know-how so the agent can "load history first, then act; ship, then persist."
The clever part is it does not build a new agent — it reuses the coding CLI you already have (Codex / Claude Code / OpenCode): `oc init` generates user-level Skills and slash commands for them, plus an MCP server so Cursor/Claude Code/Codex can call OpenContext as tools to read, write and search your knowledge base. Use it via a desktop app, a local Web UI, or the `oc` CLI, with knowledge kept local. Open source, one npm line to install the CLI.
Persistent context memory: Saves project background, decisions and knowledge that the agent can load across days, repos and sessions, no re-explaining.
Reuse your agent: Uses your existing Codex/Claude Code/OpenCode CLI, no separate agent subscription.
Global knowledge base: One contexts library shared across all projects, loaded automatically by the agent.
MCP + Skills: A built-in MCP server, user-level Skills and slash commands let the agent read, write, search and iterate your knowledge.
Multiple ways to use: A desktop app for visual management, a local Web UI, or the oc CLI for automation.
Local and open source: Knowledge stays local; open source, installed with one `npm install -g @aicontextlab/cli`.
The clever part is it does not build a new agent — it reuses the coding CLI you already have (Codex / Claude Code / OpenCode): `oc init` generates user-level Skills and slash commands for them, plus an MCP server so Cursor/Claude Code/Codex can call OpenContext as tools to read, write and search your knowledge base. Use it via a desktop app, a local Web UI, or the `oc` CLI, with knowledge kept local. Open source, one npm line to install the CLI.
Features
Persistent context memory: Saves project background, decisions and knowledge that the agent can load across days, repos and sessions, no re-explaining.
Reuse your agent: Uses your existing Codex/Claude Code/OpenCode CLI, no separate agent subscription.
Global knowledge base: One contexts library shared across all projects, loaded automatically by the agent.
MCP + Skills: A built-in MCP server, user-level Skills and slash commands let the agent read, write, search and iterate your knowledge.
Multiple ways to use: A desktop app for visual management, a local Web UI, or the oc CLI for automation.
Local and open source: Knowledge stays local; open source, installed with one `npm install -g @aicontextlab/cli`.
