ADHDev: A Self-Hosted, Open-Source Dashboard Hub for AI Coding Agents

As AI coding agents shift from experimental toys to everyday engineering tools, a new challenge emerges: how do you monitor, coordinate, and control multiple assistants that operate across different terminals, repositories, and environments? A fresh open-source repository called ADHDev (Agent Dashboard Hub) aims to answer that question with a single, self-hosted interface.

What ADHDev Is (and What the Repository Tells Us)

ADHDev, published under the GitHub handle vilmire/adhdev, is described as an open-source dashboard to “monitor & control AI coding agents from a single dashboard.” Built with TypeScript and released very recently, the project already carries the thematic tags: acp, agent, cli, dashboard, development, hub, ide, open-source, and remote-control. It has attracted early interest with 51 stars at the time of writing.

The acronym ACP appears in its topics, hinting at a likely “Agent Communication Protocol”—a standardized way for coding agents to report status, accept commands, and stream logs. While the full specification isn’t spelled out in the public repo yet, the presence of both cli and remote-control signals that ADHDev isn’t just a passive viewer. It’s positioned as a command center where developers can start, stop, and steer agents from a unified web dashboard.

Why a Unified AI Coding Agent Dashboard Matters Now

The developer ecosystem is no longer limited to a single copilot. Teams today mix GitHub Copilot for inline suggestions, SWE-Agent for autonomous issue resolution, and emerging command-line agents like OpenAI Codex CLI that run hands-free on local or remote machines. Each agent works in its own silo, often with no shared visibility layer.

This fragmentation creates real pain points:

• Observability gaps: Is the PR-writing agent stuck in a loop? Is a cloud-based agent burning credits on unproductive work?
• Permission chaos: Without a control plane, it’s hard to revoke access or time-limit an agent that’s running amok.
• Context switching: Managers and lead developers jump between terminals, IDE extensions, and cloud consoles to track what each piece of AI automation is doing.

A self-hosted hub like ADHDev addresses these frictions by centralizing agent monitoring into one dashboard that the team controls—aligning with a broader push toward MCP (Model Context Protocol)-style standardisation and vendor-neutral orchestration.

Who Should Pay Attention Right Now

• Engineering leads and founders who are evaluating how AI agents fit into a ship-quality workflow. ADHDev offers a glimpse of the orchestration layer you’ll likely need.
• Developers and DevOps operators experimenting with multiple coding agents—especially those running agents in headless CI/CD pipelines or sandboxed environments.
• Open-source tool evaluators looking for lightweight, transparent alternatives to proprietary agent management systems.

If you’re already juggling several AI-assisted coding tools and feel the lack of a cockpit view, this early-stage project is a signal worth tracking.

Practical Use Cases (Based on What’s Visible)

While full documentation is still emerging, the repository’s structure and tags suggest several concrete workflows:

1. Centralized Agent Session Control

Instead of opening multiple SSH sessions or IDE extensions, an operator could see all active agents—local and remote—on one screen and terminate, pause, or adjust parameters instantly. This fits the “remote-control” tag and would be invaluable for time-boxed autonomous runs.

2. Cost and Utilization Monitoring

With multiple agents consuming API tokens, a dashboard that aggregates request counts, tokens spent, and task completion status offers immediate ROI. Even a simple real-time view prevents a runaway agent from blowing through a monthly budget unnoticed.

3. Audit Trail for Agent Decisions

When agents generate code or modify repositories, logging their actions into a single hub simplifies post-hoc reviews. ADHDev’s “dashboard” and “cli” pairing could surface what commands were issued, which files were changed, and why—helping teams comply with change-review policies partially or fully automated by AI.

4. Multi-Agent Workflow Coordination

As multi-agent patterns emerge (one agent writes code, another reviews it, a third updates docs), having a hub to define and watch these hand-offs becomes a productivity lever. While ADHDev doesn’t yet advertise workflow orchestration, the “hub” and “agent” combo lays a foundation that could evolve in that direction.

Limitations, Risks, and What Remains Unclear

ADHDev is a newly public repository, so realistic expectations are essential:

• Early maturity: 51 stars and minimal documentation mean the tool is likely in a proof-of-concept or alpha stage. Stability, bug surface, and breaking changes are unknown.
• Agent compatibility: The project mentions ACP, but it’s unclear which real-world agents—SWE-Agent, OpenHands, or custom GPT-based tools—are supported out of the box. Until a compatibility list or protocol spec appears, integration effort is a black box.
• Security posture: Self-hosting is great for data privacy but shifts the burden of securing the dashboard, auto-rotation of credentials, and agent authentication onto the user. Remote control capabilities create a powerful attack surface if not hardened.
• Vendor-neutral vs. fragmented ecosystem: If ADHDev relies on a new or under-adopted protocol, it could become yet another island rather than a true unifier.
• No benchmarks or release claims: The repository contains no published performance numbers, SDK support, or concrete launch timeline beyond its immediate availability.

These factors don’t diminish the project’s potential; they simply frame it as one to watch rather than plug-and-play today.

How to Evaluate an AI Agent Dashboard or Orchestration Tool

Whether you’re assessing ADHDev or any similar hub, use these criteria to separate vision from production readiness:

1. Protocol and integration breadth: Does it speak a standard protocol (ACP, MCP, gRPC, REST)? How many off-the-shelf agents can connect without heavy forking?
2. Deployment model: Self-hosted vs. cloud-only. Self-hosted delivers privacy but requires ops work; cloud solutions trade control for speed.
3. Security and access controls: Look for role-based access, audit logs, and the ability to enforce policy—not just view status.
4. Extensibility: Can you add custom visualizations, hooks, or alerts? Open-source projects like ADHDev allow you to fork and customize, which is a significant advantage over closed SaaS.
5. Community and governance: Early projects thrive on responsive maintainers. Check issue discussions, contribution guides, and whether the roadmap aligns with your stack.
6. Cost transparency: A dashboard may be free but still cause hidden costs if it inefficiently polls APIs or requires heavy infrastructure. Keep an eye on resource footprint.

FAQ

What exactly is ADHDev?

ADHDev (Agent Dashboard Hub) is an open-source, self-hosted dashboard designed to monitor and control AI coding agents from one interface. It’s built in TypeScript and appears to use a command-line and browser-based approach for agent interaction.

Which AI coding agents does ADHDev support?

The repository doesn’t yet publish a confirmed list. The topic “acp” suggests a future compatibility layer for agents that adopt that communication protocol. Until specifications or plugins are released, integration details are unconfirmed.

Is ADHDev ready for production use?

All signs point to an early-stage project—very recent public release, minimal community size, and sparse documentation. It’s best suited for exploration, prototyping, and contributing right now rather than production deployment.

How does this compare to commercial agent dashboards?

Commercial tools often bundle the agent, the runtime, and the dashboard into a closed ecosystem. ADHDev’s open-source, self-hosted model gives teams full control over data and customization, but it currently lacks the polish, support guarantees, and ready-made integrations of established platforms.

Where can I learn more or try it?

The source code and basic setup instructions live on GitHub under vilmire/adhdev. Because the project is evolving quickly, checking the README and open issues is the most direct way to see current capabilities.