Lobu
6.5/10Lobu is an open-source context layer and control plane where teams run AI agents with shared, permission-aware memory across Slack, WhatsApp and MCP clients.
Data last updated November 2026 · Our methodology
Overview
Lobu is an open-source workspace and runtime for organisational AI agents. Its core is an append-only event log. Connectors for GitHub, Gmail, Linear, Stripe, webhooks or custom TypeScript code feed that log, and Lobu projects it into typed entities such as Company, Project or Customer. Every agent in an organisation then reads the same record, with permissions applied. Teams can run persistent specialist agents in Lobu itself, or connect Claude, ChatGPT, Codex, Cursor or OpenClaw through a single MCP endpoint and use only the shared memory. Automations run on a schedule, on connector events or manually. They can notify, ask for approval, or take action. Agents can be reached in Slack, Telegram, WhatsApp, Discord, Microsoft Teams, Google Chat, a web app, or a REST API and TypeScript SDK. Inference can come from OpenAI, Anthropic, Gemini, OpenRouter or any compatible endpoint. Lobu suits technical teams, especially startups and developer-led companies already using coding agents, that want shared, governed context instead of separate per-tool memory. The repository is Apache-2.0 licensed, and self-hosting is free on Docker or Kubernetes with Postgres. Lobu Cloud is billed by usage: per vCPU-minute, per GB-minute of memory and per GB-month of storage, with per-second metering and no charge while agents sit idle. The main limitation is maturity. The project began in 2025 under the name peerbot.ai and has a small maintainer base and only a few hundred GitHub stars. No independent production reviews were found, there is no documented SSO or compliance certification, and an outside analysis cautions that its default execution environment should not be treated as isolation for untrusted code.
Our Rating
Who Lobu is best for
Best software for DevelopersGood fit
Lobu is configured in TypeScript and driven by a CLI and SDK. It connects to coding agents such as Claude Code, Codex and Cursor over MCP.
Self-hosting is free and the cloud has no minimums and scales to zero, which suits small technical teams.
Connectors for GitHub, Linear and Stripe, plus shared customer entities, fit SaaS operations such as tracking customer risk.
Less suited
Docker and Helm deployment, credential brokering, egress policy and approval gates suit IT teams that operate agents, though SSO is not documented.
Lobu has governance features, but a small maintainer base, no published compliance certifications and no large support organisation make it a weaker fit for enterprises.
Pros & Cons
Pros
- Apache-2.0 licensed and free to self-host on Docker or Kubernetes, so there is no per-seat fee and no vendor lock-in.
- One MCP endpoint lets Claude, ChatGPT, Codex, Cursor and OpenClaw share the same entity-based memory, so teams can keep the agents they already use.
- Wide channel coverage: Slack, Telegram, WhatsApp, Discord, Microsoft Teams, Google Chat, a web app and a REST API.
- Governance controls include credential brokering by the gateway, ACL-filtered memory, tool policies, approval gates, egress policy and an append-only audit history.
- The cloud is metered per second and costs nothing while agents are idle.
- Model-agnostic: works with OpenAI, Anthropic, Gemini, OpenRouter or any compatible endpoint, configured per project or per agent.
Cons
- Young project with a small maintainer base and a few hundred GitHub stars, which raises bus-factor risk.
- Setup is code-first, using lobu.config.ts and a CLI, so non-technical teams will struggle without developer help.
- No SSO, SOC 2 or other compliance certifications are documented.
- An independent analysis warns that the default just-bash execution environment is not real isolation for hostile code, so a separate sandbox is needed.
- No independent user reviews on G2, Capterra or Reddit, and support appears limited to founder contact and GitHub.
- Cloud pricing is metered with no published examples, so monthly costs are harder to predict than with flat plans.
Key Features
Lobu Pricing
Self-hosted
- Full feature set
- Free under the Apache-2.0 open-source license
- Docker single-host or Kubernetes Helm deployment
- You pay for your own infrastructure and LLM API usage
- Embedded Postgres for local use, external Postgres for production
Lobu Cloud
- Idle agents cost nothing
- No minimums or reserved instances
- Managed hosting of the same codebase
- Per-second metering while agents are active
- Access arranged through a talk-to-founder contact
- Usage-based billing per vCPU-minute, GB-minute of memory and GB-month of storage
Frequently Asked Questions
Yes, if you self-host. The code is Apache-2.0 licensed and self-hosting includes all features, though you pay for your own servers and LLM API usage. The managed Lobu Cloud is billed by usage.
Lobu Cloud charges about 0.000463 USD per vCPU-minute, 0.000231 USD per GB-minute of memory and 0.15 USD per GB-month of storage. Usage is metered per second, idle agents cost nothing, and there are no minimums or reserved instances.
Yes. Lobu exposes a single MCP endpoint that Claude, ChatGPT, Codex, Cursor, OpenClaw and custom agents can connect to for shared memory, without running Lobu's own agent loop. Each external agent needs its own authorised connection.
Agents can run in Slack, Telegram, WhatsApp, Discord, Microsoft Teams, Google Chat, a web app or through a REST API. Built-in connectors cover GitHub, Gmail, Linear and Stripe, and custom connectors can be written in TypeScript and run in an isolated V8 sandbox.
Lobu can run locally with embedded Postgres, on a single host with Docker, on Kubernetes with a Helm chart and external Postgres, or as the managed Lobu Cloud. All of them use the same lobu.config.ts project definition, applied with the CLI.
The gateway authenticates requests and resolves ACLs, and it brokers credentials so workers do not see raw secrets. Lobu also applies guardrails, tool policies, approval gates, an egress proxy and an append-only event log for audit. SSO and compliance certifications are not documented, and an independent review advises pairing Lobu with a real sandbox for untrusted code.
Lobu works with OpenAI, Anthropic, Gemini and OpenRouter, plus any compatible endpoint an organisation defines. Providers are set per project, and each agent can state its own preferences.
Similar Tools
See all alternatives
LangChain offers open-source frameworks (LangChain, LangGraph, Deep Agents) plus the paid LangSmith platform. Developers use them to build, evaluate, deploy and monitor LLM agents.
Open-source platform for building LLM apps, agents, RAG knowledge pipelines and visual AI workflows, available as cloud SaaS or self-hosted.
CrewAI is an open-source Python framework and enterprise platform (AMP) for building, deploying, and governing multi-agent AI workflows. It is used by developers and enterprise platform teams.
Ranked near Lobu in AI Software
See Lobu in the listLobu is #167 in AI Software by our score.
