> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rundesert.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Bot detection
> Why some sites block your workspace, and the two ways around it.
Your workspace runs in a cloud sandbox. Every request it makes leaves from a data
center, not from your laptop.
Plenty of sites watch for exactly that. If a site runs bot protection, it will
very likely spot your workspace and block it. This is normal, it is not a bug in
your automation, and you have two ways to fix it.
## What being blocked looks like
The agent will tell you when it hits a wall, usually while it runs the job by
hand for the first time. You might also see it as:
* A captcha that never clears, no matter how many times the page reloads.
* A 403 or "access denied" response.
* A page that loads but comes back empty, or holds nothing but a "checking your
browser" message.
* A warning about unusual traffic from your network.
## Fix one: get a developer token
Try this first. It works better, costs less, and breaks less often.
Most big sites offer a developer token, sometimes called an API key or app
credentials. The site builds it for exactly this purpose: letting a program read
data without pretending to be a person in a browser.
A token beats fighting the block on every count:
* Bot protection does not apply to it, because you are no longer sneaking in.
* It runs in a fraction of the time, since your automation skips loading a whole
web page.
* It keeps working when the site redesigns, because you read clean data instead
of scraping the layout.
You find these on the site's own settings or developer page. Grab the token,
paste it into the chat, and the agent takes it from there.
### Example: Reddit
Say you want a daily digest of a subreddit. The agent opens Reddit from your
workspace, and Reddit blocks it, because it sees automated traffic from a data
center.
Reddit also hands out developer credentials to anyone with an account. Go to
**reddit.com/prefs/apps**, create an app, and Reddit gives you a client ID and a
secret. Paste both into the chat:
> Reddit is blocking you. Here are my developer credentials.
> Client ID: `xxxxxxxx`
> Secret: `yyyyyyyy`
Now the agent reads Reddit through the front door. No block, no captcha, and the
digest runs every morning without anyone touching it.
The same move works on most sites worth automating. GitHub, Stripe, Linear,
Notion, and Airtable all hand out tokens from their settings pages.
Check for a token before you start, not after you get blocked. It saves the
agent a wasted attempt, and you get a faster automation out of it either way.
See [Logins and tokens](/prompting#logins-and-tokens).
## Fix two: ask us for a managed proxy
Some sites have no developer token, and some block you anyway. That is when the
IP address itself is the problem.
We can put your workspace behind a managed proxy. Your requests then leave from a
residential IP, which looks like a normal home internet connection instead of a
data center, and most bot protection lets that traffic through.
Email **[support@rundesert.com](mailto:support@rundesert.com)** and tell us which
site you are trying to reach. A managed proxy costs extra, so we will walk you
through the pricing before we turn anything on.
Ask about this only after you have checked for a developer token. A token is
free, faster, and steadier. A managed proxy is the answer when no token exists,
not the first thing to try.
## Which one do I need?
Search the site's settings for "API", "developer", or "apps". If you find
one, take it. You are done.
Not sure whether a site has one? Ask. The agent reads the site's own
documentation and tells you where to click.
> Does this site have an API token I can get you? Where do I find it?
No token and still blocked? Write to [support@rundesert.com](mailto:support@rundesert.com) about a managed
proxy.