Emilia Protocol

LiteLLM Agent Platform

A control plane from the LiteLLM team that gives one API and gateway for calling many agents, such as OpenCode, Claude Managed Agents, Cursor Agents and Deep Agents.

Agent Barn

Agent Barn is an open-source (Apache-2.0) control plane for running a fleet of AI agents on your own Kubernetes: create, configure, run and monitor agents across Slack, Microsoft Teams and Telegram from one place.

HiRoute

HiRoute, from the Higress team, is an open-source local model router for long-running agents like Codex and Claude Code: it hands routine stages to cheaper models, saves your strongest model for hard steps and keeps a record of each decision.

LobsterMatch

LobsterMatch is a beta professional network for autonomous AI agents where operators register an agent with a public profile, match with complementary agents, coordinate tasks in signed 1:1 dialogs and sessions, and record completed collaborations on both profiles while keeping runtime credentials private.

Fileaify

Fileaify is a project workspace that plugs into the coding agent you already use, such as Claude Code, Codex, Cursor or Windsurf. You give the agent one URL (fileaify.com/agent) or the hosted MCP connector, approve its access, and it finds the project's documentation, architecture decisions, roadmap and known issues in each new conversation, so you don't have to explain the project again, even when you switch agents. Credentials stay in a vault: the agent can run approved actions with them without ever seeing the keys, sensitive actions need a human's approval, and paid tiers add a sandboxed executor where the agent can actually run work. The agent also writes its contributions back into the project docs, following your review rules. A live demo project lets you try it without installing anything.

Foreman

Foreman from ThruWire supervises coding agents with fast decision models. You give it a ticket, spec or bug report and it runs Codex, OpenCode, Hermes or LangChain's Deep Agents Code as workers, or attaches to supported sessions such as Pi, then independently judges whether the job meets the original requirements, whether the test evidence is enough and when verification or human input is needed. Extra checks can be configured for your workflow, and it is MIT licensed.