Lobu: Open-source control plane to unify organizational AI agents
lobu, from Lobu Ai, is an open-source backend and control plane that turns separate AI assistants into coordinated organizational teammates. It operates as a Model Context Protocol server that offers a shared runtime connecting language models with internal tools and messaging surfaces. The app emphasizes permission-aware access, developer-oriented connector tooling, and background automations for multi-channel deployments. Target users include engineering teams, AI platform owners, and organisations building self-hosted agent fleets who need a protocol-first, extensible runtime.
How Lobu records provenance for agent decisions
The tool uses an event-sourced approach, treating every tool call and message as an append-only event log that creates a verifiable audit trail. That design gives agents a provenance record for where facts originated, and it makes it possible to inspect the sequence of events that led to a decision. Auditability is embedded in the memory model rather than bolted on, which matters for compliance workflows and debugging agent behavior.
How it fits into developer workflows and integrations
The platform provides a TypeScript-based connector SDK and a single configuration file, lobu.config.ts, to declare identities, skills, and permissions across deployments. The tool runs on Node.js and deploys with Docker, Kubernetes, or a local CLI, and it typically requires PostgreSQL for persistence. These choices align the project with standard backend stacks used by engineering teams.
What external systems and connectors it supports
The tool ingests external activity via built-in connectors for systems such as GitHub, Gmail, Slack, Jira, and Linear, letting agents observe company events as part of the append-only log. Connectors turn external notifications and API calls into persistent events, so integrations contribute to the shared memory and can trigger versioned background tasks.
Community profile and protocol compatibility
The platform is protocol-first and functions as a Model Context Protocol server, which preserves compatibility with MCP-compatible clients and emerging agent ecosystems. The project is open source with active contributions and visibility on developer forums and code hosting platforms, which supports extension and third-party connector development.
Practical judgement on who should adopt it
The tool is a developer-focused option for teams that need a self-hosted, auditable agent runtime and can allocate engineering resources to operate backend infrastructure. It rewards organisations that prioritise protocol compatibility and data sovereignty. Teams seeking a drop-in, nontechnical solution should consider more managed alternatives; engineering-led platforms benefit most from its event-sourced design and connector tooling.
