01
Detect
See the spike.
Runs, token use, cost, errors, and health put the problem on one screen.
Observability tools already do this part well.
Control plane for live agents
Operate the agents you already shipped.
Bring your live agents into one control panel. Track every run, understand the cost, and pause, restart, swap the model, push a prompt, or teach it a new skill—without redeploying the target app.
No framework adoption · No LLM proxy · Integration fails open
Your stack. One operating view.
Built around the tools you already use.
Keep your models and your builders. Connect ACP through the SDK or HTTP integration inside your app.
Model providers
OpenAI
Anthropic
Gemini
Groq
DeepSeek
Mistral AI
Meta Llama
Ollama
OpenRouter
Cerebras
Qwen
Kimi
MiniMax AI
Z.ai / GLM
xAI
Perplexity
Hugging Face
NVIDIA
IBM
Microsoft
Meta
Thinking MachinesApp builders
Replit
Lovable
Bolt.new
Base44
Bubble
Firebase Studio
v0Agent & workflow tools
LangChain
LangGraph
CrewAI
Langflow
n8n
Make
Zapier
Lindy
Manus
Salesforce AgentforceCompatibility depends on access to your app’s code or HTTP hooks. Provider usage and accounts stay with you.
03:07 · The operating gap
From the first signal
From the first signal
to the next safe run.
When an agent loops at 3am, observability tells you what happened. ACP carries the journey through the missing half: isolate, intervene, and verify—while the app stays deployed.
02
Diagnose
Find the agent.
Separate the misbehaving agent from the rest of the app before changing anything.
No codebase hunt. No guessing which deploy is live.
03
Intervene
Pause it now.
Send the command to the integration already running inside the target app.
The local wrapper checks state before the next LLM call.
04
Verify
Watch the line flatten.
The next heartbeat confirms the new state. Production stays up; the model calls stop.
No prompt edit. No build queue. No redeploy.
The 3-minute connect
Your app. Your agents.
Your app. Your agents.
One connection.
Create an app, install agent-control-panel or paste the generated instructions into your coding agent, and wait for the first heartbeat. Scroll to follow the connection, from your first key to a live heartbeat.
How it works
Two directions, one wire.
Your app reports upward on its own schedule. ACP sends signed commands downward. Nothing sits in front of your LLM traffic, so ACP being down cannot take your agents with it.
Heartbeat every 30s
Status, tokens, runs and errors since the last beat. Run reports carry model used, latency and prompt version.
ACP holds the state
Six commands, one audit trail. Prompt pushes are approval-gated before they reach the app.
HMAC, timestamp, nonce
Signature over ts.nonce.body, a 300-second freshness window, and a replay cache that rejects a reused nonce.
See + operate
See what happened.
See what happened.
Decide what happens next.
Five surfaces, one control plane: send commands, read failing runs, account for tokens, score responses, and turn a repeated failure into a deployed skill.
01Surface 01 of 05
Control
Pause, restart, swap the model or push a prompt. The command lands in the integration inside your app.
ACP never routes your LLM traffic. Local state is checked before the next call, so nothing waits on a deploy. Pick a command to send it in the panel.
02Surface 02 of 05
Runs
Open the failing run, read the reason it failed, and get the fix in the same view.
Every run report carries model, latency, tokens, tool calls and the error. Failures are grouped so a repeated cause is obvious instead of buried in a log.
- Filter to failures, one agent, or one app.
- Read the error, the explanation of what the agent did, and the steps to fix it.
- Act from the run: retry it, push a prompt, or send it to the Skills Hub.
03Surface 03 of 05
Token usage
Account for spend per agent, per app, per model, and per day.
Input and output are counted separately, because the fix for a long prompt is not the fix for a long answer. Compare the last 7, 30 or 90 days.
- Split input against output on every agent and every day.
- Group by agent, app or model to find the one line that moved.
- Compare 7, 30 and 90 day windows against the run count behind them.
04Surface 04 of 05
Feedback
Score a response by hand, or let a reviewer model score it. Both teach the agent.
Feedback attaches to the run, not to a spreadsheet. When scores for one agent keep landing low, the same evidence becomes the case for a new skill.
- Rate any run yourself, with a note the next reviewer can read.
- Auto-score runs with a reviewer model on resolution, policy and tone.
- Escalate a pattern of low scores straight into a skill draft.
05Surface 05 of 05
Skills Hub
When an agent keeps failing, it gets taught a skill instead of a patch.
ACP drafts the skill from the failing runs, replays those runs against it in the Skills Testing Lab, and compares the agent before and after. The skill deploys only if the overall result clears the gate.
- Draft a skill from the failures that prove it is needed.
- Test it in the lab on replayed runs, before and after, side by side.
- Deploy on pass only. A skill that does not clear the gate stays a draft.
Control without a proxy.
The wrapper checks paused or stopped state before each LLM call.
traffic routeunchanged
agent stateremote
Change the next run.
The integration resolves the override locally before the request.
beforegpt-4.1-mini
afterclaude-sonnet-4-6
Fetch, cache, fall back.
Successful fetches are cached; failures fall back to cache, then hardcoded prompt.
currentv12
incomingv13 · cached
Your app keeps running.
Heartbeat and report failures are caught so ACP stays outside the critical path.
ACP unavailablecaught
agent runcontinues
Product tour
The screens you operate from.
Honest comparison
Understand every run. Act on what you find.
ACP brings run diagnostics, quality scoring, and replay evaluation together with live agent controls. Compare the workflows you need, from investigating a failure to testing a fix and applying an approved change.
| Capability | ACP | Langfuse | LangSmith | Helicone | Braintrust | Datadog |
|---|---|---|---|---|---|---|
| Run diagnostics and evaluations | Run diagnostics · feedback scoring · replay evaluations | Scores & experiments | Online & offline evals | Tracing | Scorers & experiments | APM |
| Retrofit an already-shipped app | Generated integration | Instrumentation | Instrumentation | Gateway or SDK | Instrumentation | Instrumentation |
| Pause one hardcoded agent remotely | Yes | Not its job | Not its job | Not its job | Not its job | Not its job |
| Swap model without redeploying | Local override | App-managed | App-managed | Via gateway | App-managed | App-managed |
| Push prompt without redeploying | Versioned fetch + webhook | Prompt management | Prompt Hub | Via gateway / SDK | Prompt environments | App-managed |
| Hosted maturity and ecosystem | Early | Established | Established | Established | Established | Enterprise-grade |
Swipe table to compare →
Security · fail-open
A control plane you can trust with production.
Remote control is only acceptable if the wire is signed and the failure mode is boring. Both are enforced in the generated integration, not in documentation.
HMAC-SHA256 over ts.nonce.body
Compared with a timing-safe check, and verified against the current and previous secret so rotation never drops a command.
300-second window, nonce cache
A stale timestamp is rejected with 403. A nonce that has already been seen is rejected with 403.
Reporting errors are caught
Heartbeat and report failures log and return. A prompt fetch failure falls back to cache, then to the hardcoded prompt.
No proxy, no key custody
Your provider keys stay in your app. ACP never sits between your agent and the model.
Objections
The questions a skeptical developer asks.
Answered against what the code actually does today.
Who built it
Built from the operator’s seat.
The founder designed, built, and operated the system alone for a year as the real control plane for a crypto venture fund. It managed 320 agents across 31 applications in daily production.
That history proves the operating model. ACP currently has zero external connected apps; the public product is at the beginning of that validation.