Agentrq
AgentRQ is a human-in-the-loop control plane that keeps Claude Code, Gemini CLI and other ACP agents waiting on your approval before they
AgentRQ earns its place when your problem is not agent capability but agent permission. Claude Code connects natively, everything else goes through the acp-gateway, and one workspace per agent keeps projects from bleeding context. The weak spots are real: collaboration is one shared Slack channel per workspace, there's no Jira or Linear sync, and you'd better be comfortable with MCP endpoints and .mcp.json. If you want agents that never ask, pass; if you want a checkpoint, this fits and costs
Verified 1d ago · liveness 74/100 · cite: rightaichoice.com/tools/agentrq
- Developers running Claude Code or Gemini CLI who want to approve actions from a phone
- Small teams running several ACP-compatible agents behind one dashboard
- Workflows that need a human checkpoint before a destructive command runs
- Teams running agents on their own machines who want remote terminal access from anywhere
- Anyone who wants agents that run to completion without asking a human
- Non-developers wanting a no-code automation builder
- Teams expecting two-way Jira or Linear sync rather than one Slack channel per workspace
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip AgentRQ if you want your coding agents to run to completion without ever pausing for a human decision, or if your team's approvals live in Jira or Linear rather than a Slack channel.
Self-hosting under Apache-2.0 means you carry the compute, daemon maintenance and upgrades yourself, so the $0 beta price is not the whole cost for self-hosters.
Free during beta for solo developers and small teams running coding agents, with no credit card required. That undercuts most agent-observability and orchestration tools that charge per seat or per run. For a self-hosting team, the Apache-2.0 licence replaces a subscription with your own infrastructure cost. Teams that need Jira or Linear-style approval routing should compare the tooling they already pay for before adding another dashboard.
In short
Agentrq — AgentRQ is a human-in-the-loop control plane that keeps Claude Code, Gemini CLI and other ACP agents waiting on your approval before they. Best for Developers running Claude Code or Gemini CLI who want to approve actions from a phone, Small teams running several ACP-compatible agents behind one dashboard, Workflows that need a human checkpoint before a destructive command runs. Free to use.
What's new in Agentrq
Checked 2 days agoAcross the latest 10 updates: 4 feature updates, 1 launch, 1 changelog entry, 2 community discussions and 2 news mentions.
AgentRQ Blog: The Pano — a shared board for agents in every workspace
New shared board of markdown docs across workspaces; agents post plans, contracts and hand-offs that other agents read and pick up.
AgentRQ ACP Gateway auto-discovers agents and their models
ACP Gateway finds every ACP-compatible agent a machine can run and each model it offers, without a hand-maintained list.
AgentRQ compares autonomous agents vs. agentic workflows
Post frames the two approaches by who decides the next step and who approves it, using Anthropic's framing and a selection table.
AgentRQ explains YOLO mode flags across coding agents
Post rounds up the YOLO-mode flag in Claude Code, Codex, Antigravity, Gemini CLI, Copilot, Cline and OpenCode, plus safe-use guidance.
AgentRQ publishes ACP support table covering 33 coding agents
Support table lists 33 ACP-compatible agents, including Claude Code, Codex, Gemini CLI, Cursor and OpenCode, with each one's ACP command.
AgentRQ for Chrome extension launches
Chrome extension puts Claude Code terminals in the toolbar and bridges any site's WebMCP tools to local or remote agents.
AgentRQ argues for one workspace per agent
Post makes the case for isolating each AI agent in a workspace doing one job: truer memory and smaller starting context.
AgentRQ stats chart task latency per workspace
Stats pages now chart closed-task duration per workspace and across the account, split into worked, blocked and waiting on you.
AgentRQ records a per-task timeline
Every task status change is recorded and drawn as a timeline: time worked, blocked, waiting on the user, and start to close.
AgentRQ v0.8.0 serves agent skills over MCP
Skills move off a single machine's directory and are served over MCP so every agent loads the same current playbook.
What people actually say about Agentrq — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
12 mentions across 4 sources (Hacker News, Product Hunt, Bluesky, GitHub) · researched Jul 6, 2026.
Average across the 4 sources that answered — each source counts once, not each post.
- +Bidirectional task assignment between humans and agents is genuinely novel.
- +Real-time SSE notifications under 50ms are fast and responsive.
- +Open-source (Apache-2.0) and self-hostable gives full control.
- +Free beta with no credit card required lowers adoption risk.
- +Isolated workspaces per agent enhance security and focus.
- −API inconsistency between REST and MCP cron validation causes confusion.
- −No native multi-agent routing across workspaces limits complex workflows.
- −Potential for spam abuse reported on Bluesky; no safeguards visible.
- −Limited community feedback outside HN and GitHub; support is minimal.
- −Documentation was still being written at launch time.
- • Self-hosting incurs infrastructure costs (servers, storage).
- • Future paid tiers are expected but not yet announced.
Viability Score
How well maintained and how widely used is Agentrq? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Bidirectional task assignment between humans and agents via createTask
- Isolated workspace with a dedicated MCP endpoint per agent
- Supervisor MCP orchestrating a fleet of agents from one server
- Native Claude Code <channel> push notifications through .mcp.json
- acp-gateway connects 55 ACP-compatible agents with one command
- ACP Gateway auto-discovers installed agents and the models each offers
- AgentRQ for Chrome extension puts Claude Code terminals in the toolbar
- Chrome extension bridges a site's WebMCP tools to local or remote agents
- Visual task board in list or Kanban view with Cmd+K task search
- Real-time updates over Server-Sent Events, no polling
- Pano: shared markdown board every workspace's agents read and write
- Workspace memory notes that outlive a task and load for every agent
- Task Scheduling with one-time or recurring cron tasks
- Events and Workflows chaining tasks across workspaces on a visual graph
- Tool Call History with a searchable trajectory per run
About Agentrq
AgentRQ is a real-time task manager for developers running coding agents that need a human in the loop. It works in both directions: an agent assigns a task to you, you reply, approve or reject, and the agent resumes — so the PR merge, the DB migration, or the ambiguous requirement never goes dark while you're away from the terminal. Each agent gets its own isolated workspace with a dedicated MCP endpoint, persona and mission, so several agents run in parallel without sharing context. A supervisor MCP at mcp.agentrq.com/mcp orchestrates the fleet. Connection takes about a minute. Claude Code and DeepSeek Harness connect natively (Claude Code gets <channel> push notifications through your .mcp.json); every other ACP agent goes through the open-source acp-gateway, which as of the 0.8.x line auto-discovers the agents a machine can run and the models each one offers instead of relying on a hand-maintained list. The control plane UI carries a list or Kanban task board, Tool Call History with a searchable trajectory per run, a Task Timeline splitting worked, blocked and waiting-on-you time, per-workspace memory, one-time and recurring cron scheduling, Events and Workflows that chain tasks across workspaces on a visual graph, and a Message Send Delay countdown before each outgoing message. Dictation runs through local Whisper on desktop (the phone's own dictation on mobile) and an in-browser auto-title model names tasks from their description, keeping audio on device. Pano, launched in October 2026, is a shared markdown board every workspace's agents read and write so plans and hand-offs survive a context reset. It's Apache-2.0, self-hostable, and free during beta with no credit card. The closest alternative is a per-CLI permission prompt or a plain agent dashboard like the ones shipping with Cursor and Copilot — those gate one agent on one machine. AgentRQ's bet is the fleet view plus the approval loop, and it charges you for that bet with MCP configuration work.
Behind the Verdict
Pick AgentRQ when the failure mode you're worried about is an agent doing the wrong irreversible thing, not an agent that can't do the thing at all. The approval loop — agent assigns a task, you reply from your phone, the run resumes with no re-prompting — is the whole product, and it's the part that a terminal-side permission prompt can't give you once you've walked away from the desk. The workspace model is the second reason to pick it. One agent, one MCP endpoint, one persona, one mission. Run a refactor agent and a docs agent side by side and neither one inherits the other's context. Distributed skills served over MCP (v0.8.0, September 2026) means a playbook lives in one place instead of on each machine's directory — useful the moment your agents aren't all on your laptop. Where it bites: team collaboration is one shared Slack channel per workspace, with every approval attributed but no threading, no per-project channels, and no two-way Jira or Linear sync. If your team's source of truth is a ticket tracker, AgentRQ sits beside it rather than inside it. Setup is also on you — MCP endpoints, .mcp.json for Claude, the acp-gateway command for everything else. None of that is hard for a developer; all of it is a wall for anyone else. The Machines feature deserves a mention for people running agents on hardware they own: install agentrqd and you can watch a terminal from anywhere. Combine that with Clear Context sending /clear before each task run, and you're paying for a clean slate instead of carrying the last task's tokens into the next one. Closest alternative? If you only ever run Claude Code on one machine, the built-in approval prompts plus your editor's diff view cover 80% of this. AgentRQ starts paying off around the third agent, or the first time you need
Researching Agentrq? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Agentrq actually fits — and what changes day-one when you adopt it.
You create a workspace, paste the AgentRQ entry into .mcp.json, and assign a refactor from your phone. Claude calls createTask when it needs the migration approach confirmed, you reply from the board, and the run continues without you returning to the terminal.
Outcome: The refactor lands while you are away from your desk, with the destructive step gated behind your approval rather than the agent's judgement.
You stand up one workspace per project for Claude Code, Gemini CLI and Codex CLI, connect the non-Claude agents through the acp-gateway, and watch the shared board for anything blocked or waiting on you.
Outcome: Parallel agents keep separate contexts, and the supervisor MCP lets you push new work to the right agent without re-explaining the project.
You deploy AgentRQ on your own infrastructure, install agentrqd on an internal build box, enrol it, and start an agent whose terminal you can read remotely. Distributed skills load the same playbook on every machine.
Outcome: Task and approval data stays inside your network, and you can revoke a machine from the control panel if a box is retired.
Use Cases
- Assign a Claude Code task from your phone and approve each step from the board.
- Gate a PR merge or database migration behind an explicit human approval.
- Run several coding agents in parallel with one workspace each so contexts never mix.
- Escalate an ambiguous requirement to a human instead of letting the agent guess.
- Turn on YOLO mode for read-only tool calls you already trust and keep approvals for the rest.
- Run Gemini CLI or Codex CLI through the ACP gateway with remote oversight.
- Install agentrqd on your own machine and watch the agent terminal from another device.
- Chain an event to the next task so a finished build spawns its own follow-up.
Models Under the Hood
as of 2026-10-18
Limitations
- AgentRQ is a human-in-the-loop control plane, so it is only useful if you accept a human approval checkpoint between agents and execution.
- Everything outside native Claude Code channels connects through the ACP gateway, which adds a gateway process to your setup.
- Setup involves MCP endpoints and a .mcp.json edit, which is developer work rather than point-and-click.
- Pricing is free during beta; the evidence does not describe post-beta pricing.
as of 2026-10-08
Verification history
We have re-verified Agentrq 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Agentrq tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free (Beta)
$0
Where the pricing makes sense
The company stage and team size where Agentrq's pricing actually pencils out — and where peers do it cheaper.
Free during beta for solo developers and small teams running coding agents, with no credit card required. That undercuts most agent-observability and orchestration tools that charge per seat or per run. For a self-hosting team, the Apache-2.0 licence replaces a subscription with your own infrastructure cost. Teams that need Jira or Linear-style approval routing should compare the tooling they already pay for before adding another dashboard.
Setup time & first value
How long it actually takes to get something useful out of Agentrq — broken out by persona, not the marketing-page minute.
Solo developer: about 60 seconds to sign up, create a workspace and paste the .mcp.json snippet, per the vendor's own quick start. Adding a second agent family through the acp-gateway adds one command plus its config. Self-hosters should budget an afternoon for deploying AgentRQ, installing the agentrqd daemon and enrolling machines before the first task runs.
Switching to or from Agentrq
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From manual terminal babysitting: connect Claude Code once via .mcp.json and approvals move to the web board instead of your keyboard.
- →From per-machine prompt directories: move playbooks to MCP-served distributed skills so every agent loads one current version.
- →From ad-hoc agent scripts: use Events and Workflows to chain tasks across workspaces instead of hand-launching each run.
- →From unmanaged home hardware: install agentrqd on a machine you own and enrol it for remote terminal access.
- ↗To running agents fully unattended: remove the human checkpoint, which AgentRQ is designed around.
- ↗To a Jira or Linear approval workflow: AgentRQ's collaboration surface is one shared Slack channel per workspace, so plan to rebuild routing there.
- ↗To a no-code agent platform: AgentRQ's setup is MCP config and tokens rather than a visual builder.
Integrations
Resources & Guides
- Documentationagentrq.com
Docs · Agentrq
Full product docs from agentrq.com
- Quickstartagentrq.com
Quick Start · Agentrq
Get up and running fast from agentrq.com
- Documentationagentrq.com
Claude Code · Agentrq
Full product docs from agentrq.com
- Documentationagentrq.com
Acp Gateway · Agentrq
Full product docs from agentrq.com
- Documentationagentrq.com
Workspaces · Agentrq
Full product docs from agentrq.com
- Documentationagentrq.com
Tasks · Agentrq
Full product docs from agentrq.com
- Documentationagentrq.com
Machines · Agentrq
Full product docs from agentrq.com
- Documentationagentrq.com
Workspace Cli · Agentrq
Full product docs from agentrq.com
- Documentationagentrq.com
Workspace Mcp · Agentrq
Full product docs from agentrq.com
- Documentationagentrq.com
Supervisor Mcp · Agentrq
Full product docs from agentrq.com
Tutorials & Learning
YouTube returned 6 videos for “Agentrq”, and we withheld 6: 6 could not be judged, because “Agentrq” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Agentrq.
Official links
Featured Head-to-Head Comparisons
Agentrq vs Locus Robotics
Locus Robotics and Agentrq operate in entirely different domains—physical warehouse automation vs. AI agent task management. Choose Locus if you need to boost warehouse picking productivity 2-3x with flexible AMRs and Physical AI; choose Agentrq if you're a developer wanting to manage Claude Code, Gemini, or Codex agents with real-time human oversight. No overlap in use case.
Agentrq vs Truleo
Truleo and Agentrq serve completely different verticals. Truleo is a specialized law enforcement intelligence platform connecting siloed police data (RMS, CAD, jail calls) to automate leads and report writing—ideal for detectives and command staff. Agentrq is an open-source developer tool for managing AI coding agents (Claude Code, Gemini, Codex) with real-time human oversight. Choose based on your domain: police work or software development.
Agentrq vs Presto Voice
If you run a QSR drive‑thru and want automated order‑taking with proven upsell lift, Presto Voice is your choice — but expect enterprise pricing and limited flexibility. If you’re a solo developer or team using Claude Code, Gemini, or Codex and need real‑time human oversight for agentic coding workflows, Agentrq is the free, open‑source way to get a human‑in‑the‑loop dashboard. The tools serve entirely different domains; pick based on whether your bottleneck is drive‑thru operations or AI agent management.
Popular in Agent Frameworks & Orchestration
Temporal AI
Durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned sessions.
Frequently Asked Questions
Used Agentrq? Help shape our editorial sentiment research.