Check your sites are ready for AI agents
Agent readiness checks whether AI agents can find, read, and use your product from your website and docs site. It reads both sites the way Claude, ChatGPT, and Perplexity fetch them, grades what it collected against 23 checks, and gives you a score out of 100, your Blockers and Gaps, and a ranked list of fixes.
It is for developer-tool and API companies who want agents to recommend their product and build with it. For what each check looks for and how the score adds up, see Agent readiness checks and scoring. To try it without an account, enter your sites at ekline.io/agent-readiness.
Before you begin
Section titled “Before you begin”- Your website and docs site are public. The check reads only what anyone on the internet can reach.
- To hand a fix to the Docs Agent, connect a docs repository to EkLine.
Run your first check
Section titled “Run your first check”-
Click Agent readiness in the sidebar.
-
Under Run your first check, confirm the Website and Docs site addresses. EkLine fills them in from your Company Profile. Docs site is optional: leave it empty to check the website alone.
-
Click Run check.
The page shows each step as it runs, with the number of pages read so far and the time left. A check takes about 4 minutes, and you can leave the page while it runs.
If a site can’t be checked, the field says why before the check starts:
| Message | What to do |
|---|---|
| We can only check public sites or does not resolve | Enter the public address of the site. |
| moved to another address | Click Use with the new address to check that instead. |
| sends visitors to a login | Click Change the address to enter a public one, or My docs are private on purpose to check what agents can reach without signing in. |
Read the report
Section titled “Read the report”The report has three tabs.
Overview
Section titled “Overview”- Score out of 100, with its band: Ready, Partial or Blocked. While any Blocker is open, the score stays at 49 or below.
- Blockers stop agents outright. Gaps make agents miss or misread something. Polish is worth doing once the rest passes.
- Fixes done counts the fixes that now pass.
- Start here is the one fix to make first. All fixes lists every fix by rank, with its level, Who fixes it, Effort and status.
- Readiness by stage shows Find, Read and Act for your website and for your docs and API, each Ready, Partial or Blocked.
All checks
Section titled “All checks”Every check with its result on each site. Filter by Apply to you, Failing, Passed, Don’t apply or Tracked. Click a check to see why it passed or failed, why it counts as a Blocker, Gap, or Polish, and Pages we read: each address the check fetched and the response it got.
History
Section titled “History”Each check you ran, with its score, Blockers and Gaps, and what changed since the one before. Click View on a row to open that check’s report as it was then.
Confirm the answers about your product
Section titled “Confirm the answers about your product”Some checks depend on what your product offers. After your first check, the report asks five yes-or-no questions, such as whether developers can create an API key on their own and whether your docs should need a sign-in. Each answer shows whether the check found it on your site or assumed No because it couldn’t tell.
Change any answer that is wrong, then click Confirm and update score. The score updates without a new check. To change an answer later, click Change next to Answers about your company under the report title.
Fix what fails
Section titled “Fix what fails”Click Open fix on any fix to see what’s wrong, how to fix it, which pages it affects, and when it counts as done. What you can do from there depends on who fixes it:
| Who fixes it | What to click |
|---|---|
| Docs Agent or Marketing Agent | Start opens a session of that agent with the fix. When it has a pull request, the fix shows Review PR. |
| Your engineers | Copy prompt for your coding agent, then paste the prompt into the coding agent your engineers use. |
Starting a fix starts an agent session, which uses credits like any other session. If no docs repository is connected, Start asks you to connect one, or to copy the prompt instead.
After the change is live, click Check again on the fix. The fix shows Passed, or Still failing with what changed. To check everything again, click Check again at the top of the report.
If a firewall blocks the check
Section titled “If a firewall blocks the check”If bot protection answers every request with a challenge page, even a browser’s, the report says We couldn’t check the site and shows the other site’s results. Click How to let checks through on that notice. Add a rule in your firewall or bot protection that skips the challenge for EkLine’s user agent and the addresses listed at ekline.io/bot. Then, click Check docs again or Check website again. AI agents most likely get the same challenge, so this is also your first fix.
Change the sites it checks
Section titled “Change the sites it checks”Click Edit sites at the top of the report. Besides Website and Docs site, you can give the addresses of your OpenAPI file, MCP server, SDK docs and CLI docs.
The checks about an API, MCP server, SDK, or CLI run only for what the last check found or what you enter here. Add the ones the check missed, or clear a link to check without it. Changes apply from the next check.
Share a report
Section titled “Share a report”Click Share at the top of the report.
- Workspace link opens the live report for members of your organization.
- Public read-only link, once you turn it on, opens the report for anyone, without fix status or agent sessions. Turn it off to stop sharing. You can share a finished full check, not a Check again on one fix.
Your latest score, Blockers, and Gaps also show on the Dashboard.
Related
Section titled “Related”Built on Potluck Docs, our open source template.