Skip to content

My agent is slow or unresponsive

“My agent is slow” usually means one of four things. Let’s narrow it down.

1. First response was instant, now it’s slow

Section titled “1. First response was instant, now it’s slow”

Most likely cause: your agent is in the middle of a long-running task (deep research, heavy sub-agent delegation, big file processing).

How to check:

  • Dashboard → Chat. Is there a “thinking” or “running tool” indicator?
  • Dashboard → Tasks. Is there an active task with an estimated completion time?

What to do:

  • Wait — most tasks complete within 30 seconds. Deep research can take 2-5 minutes.
  • Check tool output — your agent usually shows “searching…” / “browsing…” / “analyzing…” as it works. If one of those has been stuck for 3+ minutes, the underlying tool may be hung.
  • Cancel and retry — click the stop button next to the response in the chat. Then try again with a simpler prompt.

2. Everything is slow — even short messages

Section titled “2. Everything is slow — even short messages”

Most likely cause: the LLM provider is slow. MiniMax and other providers occasionally have degraded performance, and we can’t make them faster from our side.

How to check:

  • Status page — we track provider health and post incidents here
  • Try asking your agent a trivial question (“What’s 2+2?”). If even this is slow, it’s a provider issue.

What to do:

  • Wait it out — provider slowdowns usually resolve in 5-30 minutes
  • Switch providers via BYOK — if you’ve added your own API key (Anthropic, OpenAI, MiniMax, etc.) from Settings → Agent → Bring Your Own Key, your agent can route requests through it instead of our shared pool. This is the most reliable workaround for provider slowdowns, and BYOK works on every tier.
  • Report it — if it’s been slow for an hour+, email support so we can verify

Most likely cause: your agent’s pod crashed or was restarted.

How to check:

  • Dashboard header — is there an “Agent offline” warning?
  • Open Dashboard (the main landing page) — the agent status block at the top shows the agent ID, last heartbeat, and health.

What to do:

  • Wait 60 seconds — if a pod restart is in progress, it should self-recover
  • Force a restart from the Dashboard — find the status block at the top and click Restart. The button confirms with a 30-second estimate before anything happens.
  • Check the status page — sometimes we’re rolling out updates
  • If still offline after 5 minutes, email support

4. Messages are sent but the agent seems confused

Section titled “4. Messages are sent but the agent seems confused”

Most likely cause: context compression or a memory issue.

If your agent is giving answers that miss earlier context or contradict itself, the session may have been compressed aggressively. Compression is a feature (it keeps long conversations affordable) but sometimes it drops important details.

What to do:

  • Start a new conversation — click the New chat button at the top-right of the chat window. Your agent loads fresh with its long-term memory intact; only the running conversation thread resets.
  • Remind it of key context in the first message — “As a reminder: I’m working on X, and we established earlier that Y.”
  • Check MEMORY.md from the Memory tab. If something important isn’t there, add it.

status.myagentplatform.com shows:

  • Current platform health (LLM providers, agent nodes, email/calendar integrations)
  • Active incidents
  • Scheduled maintenance windows
  • Recent incident history

If you see an incident posted that matches what you’re experiencing, we’re already working on it. No need to email — updates will be posted to the same status page.

If you email support about performance issues, include:

  • Your agent ID (shown at the top of your Dashboard page next to the status block)
  • The approximate time the issue started
  • What you were doing when it started (sent a message, triggered a scheduled task, etc.)

This lets us pull the relevant pod logs and diagnose faster. We look at logs only for the specific agent/time you identify, and only long enough to understand the issue.

Clicking Restart on the Dashboard status block is a safe operation. It:

  • Gracefully shuts down your agent pod
  • Starts a fresh pod
  • Reloads MEMORY.md and USER.md
  • Keeps your workspace and scheduled tasks intact

A restart takes 30-60 seconds. It’s a good first response to any weird behavior.