Knowledge limits
Your cap counts extracted text, not file size. Free 400 KB through Agency 60 MB, per chatbot.
Browse topics
What it is
Every chatbot has a ceiling on how much knowledge it can hold. The ceiling is set by your plan, it is measured in extracted text, and it is shared by every way of adding material — pasted text, uploaded files, website crawls, and cloud imports all draw from the same budget.
When the budget runs out, Add is blocked and crawls stop early. Neither of those is a fault; both are the cap doing its job.
The caps
| Plan | Knowledge per chatbot |
|---|---|
| Free | 400 KB |
| Starter | 10 MB |
| Standard | 20 MB |
| Pro | 40 MB |
| Agency | 60 MB |
The cap is per chatbot, not per workspace. Two chatbots on Standard have 20 MB each, not 20 MB between them. Extra knowledge add-ons, where offered, top up a shared pool the owner can draw on.
The thing everyone gets wrong
The cap counts text, not file size.
A 6 MB PDF that is mostly photographs might extract to 20 KB of text and barely touch your budget. A 300 KB plain-text export of a support inbox is 300 KB of text and will consume three quarters of the entire Free allowance on its own.
This cuts both ways, and it explains the two complaints people arrive with:
- "My file is tiny but I am out of space." The file is small; the text in it is not, or you were already nearly full.
- "I uploaded 40 MB of PDFs on Starter and it worked." Most of those megabytes were images, which extract to nothing.
If you want to predict the cost of a document, look at how much writing is in it, not at its size on disk.
When you would use it
Read this when Add shows an upgrade prompt, when a crawl finishes with far fewer pages than you asked for, or when you are planning what to put on a chatbot before you build it.
Where to find it
Open Dashboard → Chatbots → your chatbot → Knowledge and read the usage meter at the top. The same figure appears inside the Add drawer, alongside an estimate of what the material you are about to add will cost.
Steps
- Open Knowledge and read the usage meter.
- Before adding anything substantial, open Add and check the estimate against the space you have left.
- If you are close to full, clear space first — see the pruning list below.
- To raise the cap, the workspace owner opens Settings → Billing and upgrades, or buys extra knowledge if that is offered on your plan.
- Re-run whatever was blocked. A crawl that stopped at the cap has to be started again; it does not resume on its own.
What you will see
The meter is on the Knowledge page and repeated inside the Add drawer. Hitting the ceiling produces an explicit limit dialog rather than a silent failure or a mysterious error.
For a scanned document there is one wrinkle worth knowing: the size cannot be known until the page has been transcribed, so a scan is checked against your cap after transcription rather than at upload. A scan can therefore be accepted, processed, and only then rejected for size. See Scanned documents and images.
The other limits, which are not this one
Three separate sets of numbers get confused with the plan cap. They are independent, and any of them can stop you while the others have room:
Upload transfer ceilings. 10 MB per file, 20 files per batch, 90 MB per batch. These govern the transfer itself, not your allowance. See Upload files.
Cloud import ceilings. The same three numbers apply to a Drive or Dropbox import. See Import from Google Drive or Dropbox.
Crawl ceilings. Depth, page count, delay, and sitemap URL count are limits on the crawl, not on your knowledge. A crawl can be well under 100 pages and still halt because the knowledge budget ran out. See Crawl limits.
How to free space without losing answers
Work down this list; the early items cost you nothing.
- Remove duplicates. The same policy uploaded as a PDF and pasted as text is paid for twice, and worse, gives the chatbot two versions to choose between.
- Remove superseded versions. Last year's price list is not just wasted space, it is a source of wrong answers.
- Remove crawled noise. A site-wide crawl usually collects tag pages, author archives, and pagination that contain no useful answers. Delete those rows and re-crawl with exclude patterns next time — see Robots.txt and blocked pages.
- Replace bloated documents with text. A 40-page designed brochure whose actual content is two pages of facts can be retyped as text knowledge at a fraction of the cost, and it will answer better.
- Split across chatbots. If you genuinely have more material than one chatbot's cap allows, and it divides cleanly by subject or by audience, two chatbots each get their own full allowance.
Removal takes a moment to complete — the freed space appears once the removing badge clears. See Remove knowledge.
Limits and plan notes
- Message credits and knowledge size are unrelated. Running out of credits stops the chatbot from answering; running out of knowledge space stops you from adding material. Neither causes the other.
- Training does not consume credits.
- The cap applies to the chatbot, so switching plans changes the ceiling for every chatbot in the workspace at once.
- Downgrading to a plan whose cap is below your current usage does not delete anything, but you will not be able to add more until you are back under the new ceiling.
Common problems
"The file is 200 KB but I am out of space."
The cap counts extracted text and you were probably already near the ceiling. On Free, 400 KB is roughly a hundred pages of plain writing — it fills faster than people expect. Read the meter, prune, or upgrade.
A crawl added only some pages.
It stopped when the knowledge budget ran out. That is the designed behaviour: better a partial, trained crawl than a failed one. Free space or raise the cap, then crawl again with tighter include patterns so the budget goes on pages that matter.
The meter did not drop after I deleted things.
Removal runs in the background. Wait for the removing badge to clear on every row, then refresh.
I upgraded and Add is still blocked.
Reload the dashboard so the new entitlement is picked up. If it persists, confirm you upgraded the workspace that owns this chatbot — switch accounts if you are working in someone else's workspace.
I am at the cap and everything on the list matters.
Then you have outgrown the plan, and pruning further will cost you answers. Upgrade, buy extra knowledge, or split the material across two chatbots by audience.
Common questions
Why is my small file using so much of the cap?
The cap measures extracted text, not bytes on disk. A text-dense document costs far more than a large PDF that is mostly photographs.
Is the cap per chatbot or per workspace?
Per chatbot. Two chatbots on Standard have 20 MB each. Extra knowledge add-ons top up a shared pool the workspace owner can draw on.
My crawl stopped early. Is that a bug?
No. A crawl halts when the knowledge budget runs out, so you get a partial trained crawl rather than a failure. Free space or narrow the crawl, then run it again.
Is running out of knowledge space the same as running out of credits?
No. Credits stop the chatbot answering visitors; the knowledge cap stops you adding material. Neither one causes the other.
What happens if I downgrade below my current usage?
Nothing is deleted, but you cannot add more knowledge until you are back under the new ceiling. Prune sources or upgrade again.
Was this article helpful?
Related articles
Knowledge overview
Your chatbot only answers from the knowledge you give it. Here is what counts as knowledge and how training works.
Remove knowledge
Delete a source from this chatbot and get the space back. There is no undo, and cleanup is not instant.
Upload files
Add PDFs, Word files, spreadsheets, and images from your device. Ten megabytes per file, twenty files per batch.
Crawl limits
Depth, pages, delay, and parallelism are settings you choose. Your knowledge cap is separate and can stop a crawl first.
Ready to try it on your own content?
Create a free workspace, add a document, and ask the questions your team is tired of answering.