OPEN-SOURCE DESKTOP TOOL
AgentKib
A local control center for coding-agent context, instructions, MCP, and handoffs.
AgentKib helps teams and solo builders see what Codex, Claude Code, Cursor, and other coding agents can actually load in a project, diagnose drift, review changes, and create redacted handoff packages without a cloud account.
What AgentKib helps you manage
Workspace context
Discover local projects and inspect the instructions, skills, MCP connections, and effective load order each agent can use.
Diagnostics and sync
Find missing, drifted, invalid, or duplicated configuration and review a ChangeSet plus complete diff before writing anything.
Session handoff
Turn redacted Codex or Claude Code context into editable Markdown or JSON so work can continue in another agent.
Local operating tools
Manage MCP servers, browse Git read-only, and view quota, token usage, activity heatmaps, and achievements from one desktop app.
Designed for
Multi-agent projects
Useful when one repository is touched by Codex, Claude Code, Cursor, and other tools that each read context differently.
Configuration cleanup
Use diagnostics to see which rules, skills, and MCP settings are missing, duplicated, stale, or out of sync.
Changing agents midstream
Create a reviewed, redacted handoff package instead of pasting raw transcripts or losing project context.
What this product demonstrates
Local-first trust boundary
Ordinary discovery, diagnostics, and handoffs run locally without an AgentKib account, cloud database, or model API.
Review before write
The product treats agent configuration as real project state, so writes are preceded by diffs, approvals, backups, and verification.
Cross-platform desktop delivery
AgentKib is built with Tauri and Rust, with preview packages and validation work across macOS, Windows, and Linux.
Local and private by default
AgentKib's preview documentation explains its local data boundaries, redaction behavior, unsigned installers, and release status. Review the repository before using preview builds.