Connect CTlogs to Claude, Cursor or any client that speaks the Model Context Protocol, and your assistant stops guessing about hostnames and certificates. It looks them up, mid-conversation, in the same index the rest of this site runs on.
Claude · Claude Code · Cursor · any Streamable HTTP client
Each tool turns a question your assistant could only guess at into one it can answer from the record. Ask in plain words; the assistant picks the tool.
lookup_certificates
A certificate historian
Hand it a hostname or a SHA-256 fingerprint and it reads the certificate history: issuer, validity window, type, and the other names that travelled on the same certificate.
“Which authorities have issued for api.example.com this year, and when does the newest certificate expire?”
lookup_certificates
A fingerprint resolver
A fingerprint from a log line, an alert or a pinning error becomes the certificate behind it and every hostname it covers, so one reported host expands into the whole set.
“What is this fingerprint, and which hostnames share the certificate?”
index_status
A source that says how fresh it is
How large the index is and when it was last updated, so the assistant can qualify an answer instead of presenting stale data as current.
“How current is the record you are reading from?”
account_quota
An operator that knows its budget
How much of your allowance is left and when it resets. The assistant can pace a long task instead of running into the limit halfway through.
“How many lookups do I have left before you start?”
Add the server
Paste https://mcp.ctlogs.io/mcp into your client’s connector settings, as a remote or HTTP connector.
Sign in
The client opens CTlogs, you sign in and approve the connection once. Nothing is stored in a configuration file.
Ask
“Which authority issued for this host, and when does the newest certificate expire?” “What is this fingerprint?”
claude mcp add --transport http ctlogs https://mcp.ctlogs.io/mcpOther clients word it differently, but they all want the same thing: the server address above, added as a remote or HTTP connector. The sign-in step is the same everywhere.
Sign in, do not paste
The client is authorised through your CTlogs account. It never sees your API key, so no secret ends up sitting in a configuration file on a laptop.
Connect as many as you like
A laptop, an editor and a server can all be connected at once. They act as the same account, and usage is counted once against the allowance you already have.
Disconnect one, keep the rest
Each connection stands on its own. Cut off a machine you no longer use and everything else keeps working.
Sized for a conversation
Twenty-five rows by default, a hundred at most. The reader is a context window, not a data pipeline; for bulk work, query the API directly.
You need a CTlogs account with an active plan. See what that includes, or sign in if you already have one. Every tool is read-only: nothing the assistant can call changes anything.
The Model Context Protocol is an open standard for connecting AI assistants to outside tools and data. A client that speaks it can call a tool the moment a conversation needs it, instead of you moving results back and forth by hand.
No. You sign in from the client and approve the connection once.
Yes, as many as you like, all as the same account. Usage is counted once.
No. Every tool is read-only.
The same index the rest of the site uses. New certificates are ingested continuously as they appear in the logs.