Troubleshooting: Errors & Rate Limits
Resolve common chat and agent errors, including rate limiting and budget exhaustion.
Most assistant errors are temporary and clear on a retry. This page explains the common ones, what they mean, and what to do.
"Too many AI requests. Try again in Ns."#
You've sent requests faster than the per-minute limit allows. The message includes how many seconds to wait.
- What to do: wait the indicated number of seconds, then send again. The conversation stays open.
- Why it happens: there's a per-minute cap on AI requests (your firm configures the exact rate). Rapid-fire sends — or several people on the same network at once — can trip it.
Repeatedly hitting the limit can temporarily pause AI for your account as an abuse safeguard. If you see a message about requests being paused due to repeated limit violations, wait out the cooldown it gives you and slow down.
"You've used your monthly AI assistant allowance"#
Your monthly base usage is exhausted. This is a usage-budget stop, not a bug.
- Firm seat: ask your firm admin to enable additional usage for your seat, or to top up the firm's usage pool.
- Solo individual: purchase additional usage from your account settings.
- Wait it out: your base allowance resets monthly on your firm's subscription anniversary.
See AI Budget & Usage for the full picture, including how the firm-wide extra pool works.
"Session token limit reached. Please start a new conversation."#
A single conversation has grown large enough to hit its per-session cap.
- What to do: start a new chat. A fresh conversation resets the limit and is also cheaper to run.
"The engineer assistant hit a snag"#
Something failed mid-turn — for example a transient model error or a tool/connection hiccup. The error is logged automatically.
- What to do: try again. These are usually transient.
- If it persists: start a new chat, or simplify the request and retry. Repeated failures are captured for your firm's administrators to review.
The assistant didn't reply, or replied "I couldn't process your request"#
Occasionally a turn completes without a usable answer.
- What to do: rephrase and resend. Being specific about the project and the action you want ("update the beam calc in project X to a 30 psf live load") helps.
AI features look disabled#
If the assistant or its AI-powered inputs are turned off, AI may be intentionally disabled at one of these levels:
- Globally — a platform-wide pause.
- Your account — AI is switched off for your user.
- Your firm or plan — your firm doesn't have AI access enabled.
If you expect AI to be available and it isn't, check with your firm admin first — the most common cause is the firm-level setting.
Errors over email#
When you reach the assistant by email, the same limits apply. If a usage limit or internal error stops the run, you'll get an email reply explaining it rather than silence — except for the security rejections (wrong sender, no firm, no AI access), which are intentionally silent.
Still stuck?#
- Confirm your usage isn't exhausted — see AI Budget & Usage.
- Start a fresh conversation from Chat History.
- If a problem is reproducible, your firm administrators can review the logged error details for failed turns.