Docs / Reference
Limits and quotas
Beta. During the public beta, every plan has the same high limits. The numbers below will change when paid plans launch. Sections marked TODO (owner) are placeholders.
Your message limit#
Every plan has a daily and a monthly message limit. You’re stopped when you reach either one.
| Plan | Messages per day | Messages per month | Web browses per day | Web browses per month |
|---|---|---|---|---|
| Guest | 10,000 | 10,000 | 10,000 | 10,000 |
| Free | 10,000 | 10,000 | 10,000 | 10,000 |
| Verified (free, with a confirmed email) | 10,000 | 10,000 | 10,000 | 10,000 |
| Pro | 10,000 | 10,000 | 10,000 | 10,000 |
| Max | 10,000 | 10,000 | 10,000 | 10,000 |
TODO (owner) final limits and prices for each plan after the beta.
What counts as one message? Each request Shard sends to the AI model. One question can use several:
- every step of a task where Shard uses tools and then continues counts as one;
- every compaction counts as one;
- every request an agent makes counts as one.
This applies to every provider: your own API key, Shard-hosted models, Codex and Claude Code.
Where to see it: the Chat sidebar shows “Message limit: used/limit” for whichever limit you’ll reach first. Settings → Account shows both, as Daily Limit and Monthly Limit bars.
When it resets: each window starts with your first message in it. The daily window lasts 24 hours and the monthly window lasts 30 days. TODO (owner) confirm these window lengths (they’re set in the database, not in the code I could read).
Guests are counted per network (your internet connection), not per computer. Signed-in users are counted per account.
When you reach the limit, you see “You’ve hit your message limit.” Wait for the window to reset, or TODO (owner) upgrade.
Web browsing#
The BrowseWeb tool’s Search and Scrape count toward a separate web-browse limit (see the table). When you reach it, the BrowseWeb row in the chat says “You’ve hit your daily web-browse limit” (or monthly). Shard can still look things up in the official Roblox docs, which have their own short-term limit (30 lookups per minute per network).
Shard-hosted models#
Hosted models also use credits and have daily budgets. See Shard-hosted models and credits.
Other limits#
| Limit | Value |
|---|---|
| Accounts created per network | 2, ever. After that: “This network has reached its account limit.” |
| Sign-in and sign-up attempts | 10 per minute per network |
| Account changes (username, email, password) | 5 per minute per network |
| Changing your username | once every 30 days |
| Support messages | 10 per minute per network, up to 4,000 characters each |
| Time for one AI reply | about 2 minutes. Longer replies fail with “The AI took too long to respond.” |
| Connected accounts (Codex, Claude Code) | 2 requests at the same time per provider |
| RunCode time limit | 120 seconds by default; Shard can ask for more |
| Image attachments | not supported yet |
Going over a per-minute limit shows “Too many requests. Slow down and try again shortly.”
Limits that come from your provider#
When you use your own API key or a connected account, the provider’s own limits and billing also apply. Shard can’t raise them. Errors like “Your Claude account has reached a usage limit” or a provider’s rate-limit message come from the provider, not from Shard.