Changelog
What changed, and when
Changes you can notice as a user, newest first. Fixes that change nothing visible are left out.
-
- A judged alert leaves To review (Undo brings it back). In To review, Relevant and Not relevant, an alert you judge so it no longer belongs to that tab collapses away, the next alert takes the focus, and the count goes down; All keeps every alert. The IntelMCP plugin for Claude Code is now version 0.6.1.
- Photos in your alerts. When a post that matched one of your rules has photos, the alert shows them, blurred until you click "Show image". Photos are cleaned of hidden data such as location, kept 90 days in a private store, then deleted. This includes posts from the private channels and groups the operator has approved. Claude can look at them with the
post_imagestool when you ask about the evidence. - Build your watch. A setup builder on this site asks what you want to catch, where, which threat actors and sectors, and how many alerts a day, one question at a time, then writes a message to paste into Claude. It runs in your browser and sends nothing. Claude takes the answers as given: it asks only for what you said you would name, and keeps your rules near the number of alerts a day you picked.
- A simpler welcome page after you subscribe: add IntelMCP to Claude from its directory listing, tell Claude what to watch, and keep your alerts one click away in the Command Centre.
- Private channels and groups the operator has reviewed. Besides public channels, IntelMCP now collects some private channels and groups its collection account is already a member of. The operator reviews each one first: a scan of recent posts that stores no posts (it keeps counts, flags and at most two masked excerpts for the operator, deleted when a chat is declined), and only approved ones are collected. A declined one keeps only its title and kind, the scan's verdict, any tags, the decision and the operator's note, and when it was found, scanned and decided, with the internal dates of the scan's email and of its collection. If it was being collected, its channel entry stays, marked excluded, with its posts, matches and indicators deleted. In results and alerts a post from a private channel or group shows "Private channel" in place of a link.
- Webhooks: when
privateis true thelinkfield is null, because a t.me link to a private channel opens only for its members.linkis a string whenprivateis false. - Only public channels can be requested.
- The IntelMCP plugin for Claude Code is now version 0.6.0: the Command Centre shows the photos in your alerts, blurred until you click "Show image", and its description and sources view no longer say the collection is public only.
-
- Simpler triage in the Command Centre. Each alert has Relevant (with a severity menu) and Not relevant; every choice saves at once with Undo, and nothing waits for an Apply. Tabs for To review, Relevant (most severe first), Not relevant and All; other filters sit behind one Filters button. After Not relevant, "Fewer like this?" offers to hide similar alerts.
- Channel requests are vetted within minutes. Channels that cannot be found or are not public are declined with the reason; clearly suitable public channels are approved and joined; anything else is reviewed by the operator. Each request shows a short note on what happened.
- A verdict keeps the notes already on an alert, and an acknowledgement can be undone.
- A redesigned Command Centre. The Overview leads with what changed since the previous period and what needs attention, then the chart and the latest alerts in compact rows. Rules and exclusions are tables you can sort and switch; Search is one bar; Account shows your usage against the plan.
- Alert cards name the target: the target country now comes from flags and the victim's domain even when the post also names the country in words, a target appears once however it is written, and Telegram, GitHub and similar links are listed as services, not victims.
- One card per post: a post matched by several of your rules is one alert that names every rule; a verdict applies to all of them.
- Attack type on each card when the post shows it: DDoS, defacement, leak, ransomware, ICS/OT, access for sale or doxxing, from the post's words and its proof links. A forward says where it came from, and your organization's name is in the header.
- On the Investigate plan, the Command Centre opens on Search and says what the Monitor-only views would add, instead of leading to empty pages.
- A date as the end of a search includes that whole day.
- The Claude Code plugin (0.2.0) checks the Command Centre page against checksums published on this site before it creates or updates yours.
-
- The Command Centre. The IntelMCP dashboard as your own private page in Claude, listed under Artifacts as "IntelMCP Command Centre" and pinnable to the sidebar. It calls IntelMCP through your own connection, so it shows only your organization's data. Create it from Claude Code with the plugin's new
/intelmcp:command-centrecommand (plugin 0.1.8). See Command Centre. - An interactive dashboard. In claude.ai and Claude Desktop chat, asking for your IntelMCP dashboard opens a panel in the conversation: overview, alerts, rules, tuning, delivery, sources, search and account, with keyboard triage and changes you review before they are sent.
- Exclusions. Hide noisy topics, countries, channels or phrases from your alerts across all your rules, without deleting anything. A country exclusion hides an alert only when every country it names is excluded.
- Alert cards carry the post: its text with the matched word highlighted, its indicators and the target country. Click a card to open the whole post in place. A post aligns by the script it is mostly written in, so Arabic, Persian and Hebrew posts read right to left even when signed in English.
- Not every link is a victim. Uptime checkers, defacement mirrors, archives, lookup and paste sites, file shares, social networks, shorteners and .onion sites are labelled as proof or services, never as victims, and the dashboard guide leaves them out of top indicators.
- The overview chart has a column for every day of the window, quiet days included.
- New tools: your account's status, your webhooks listed, tested, deleted or given a new secret, a check of whether a source is collected, and your channel requests.
- The Command Centre. The IntelMCP dashboard as your own private page in Claude, listed under Artifacts as "IntelMCP Command Centre" and pinnable to the sidebar. It calls IntelMCP through your own connection, so it shows only your organization's data. Create it from Claude Code with the plugin's new
-
- One alert per event. Reposts and copies of the same event within 72 hours are grouped into an incident, and matches are marked new, update or repeat. An update is a later post that adds something: an identifying indicator or a country new to the event, a step up in exploitation (proof of concept, then exploited), a reply in the same channel, a channel posting it a second time, or a match for another of your rules. Exploitation terms are recognised in English, Russian, Ukrainian, Arabic, Persian and Hebrew.
- Webhooks skip repeats, carry Standard Webhooks signatures beside
X-TI-Signature, and say what matched, when it was posted and how an update joined its incident. Failed sends are retried for about an hour; a destination that keeps failing is paused. Only public HTTPS targets are accepted. See webhooks. - A shorter setup. Set up monitoring asks two questions, tests draft rules against recent history with their expected alerts per day, and creates everything with one approval, including matches from the last 7 days.
- Four guides. The connector menu lists Set up monitoring, Investigate, Dashboard and Triage matches, and the Claude Code plugin has all four as commands:
/intelmcp:setup,/intelmcp:triage,/intelmcp:dashboardand/intelmcp:investigate. - Setup checks the volume: drafts that would bring more alerts than you can review are narrowed first, the newest matches are shown after saving, and running it again tunes what you have.
- Matching: Arabic and Hebrew word rules match spelling variants and attached prefixes (
=wordleaves out the article); domain rules cover subdomains and email addresses at the domain; countries are recognised from flag emojis and country-code domains. - Channel requests say when a channel is already collected, and then no request is filed.
- Matches show who else reported it: how many other posts carried the same event, and from which channels.
- Forward tracing now finds the earliest post with the same wording, not only the chain of forwards.
- Search: quoted phrases and
-wordhold in the default search mode too. - Rules created with history count those matches in their hit counts.
- When a request asks for more results or a longer window than the plan allows, IntelMCP says what it used instead.
- Subscriptions through Polar. If you sign in without a subscription, IntelMCP replies with a checkout link; if a subscription ended, the message says what happens to its data.
-
- IntelMCP opens as a hosted MCP server at
mcp.intelmcp.io: search and monitor public Telegram channels from Claude. - Sign in with Google or a one-time code sent to your email.
- The setup interview and the dashboard as prompts, and a Claude Code plugin with
/intelmcp:setupand/intelmcp:dashboard. - New rules can check recent history when they are created.
- Only public chats are collected; messages deleted at the source are removed.
- Account data is kept for 90 days after a subscription ends, then deleted; backups are encrypted.
- This website, with the privacy policy, terms and refund policy.
- IntelMCP opens as a hosted MCP server at
-
- Collection begins. The searchable history starts here.