TL;DR
- Connected to SignalRaven, ChatGPT or Claude can answer questions across many signals at once: which companies keep appearing, what's happening around your competitors, which sources are working.
- SignalRaven connects to ChatGPT, Claude and other assistants through MCP. You add one URL, sign in and approve access.
- The assistant reads your workspace's signals, sources, ICP, watchlist and Intelligence reports. It never acts on LinkedIn or uses your LinkedIn account.
- Connecting grants reading and writing together. Running a new Intelligence report or monitoring a person spends credits, and the assistant can create or change destinations.
- To sync signals into a warehouse or an internal dashboard, use the REST API.
The most useful thing an AI assistant can do with SignalRaven is answer questions about your own data.
The Signals page shows one signal at a time, and the Companies view groups them by account. Other questions cut across signals, sources and reports: whether one competitor's page draws more of your market than another, which topics keep coming up, what changed this week. Connect SignalRaven to ChatGPT or Claude and you can ask them in plain language.
What you can ask
Here's the kind of exchange that works well. The companies are fictional.
You: Which companies appeared more than once in our signals this week?
Assistant: Three did. Northwind Freight had three signals: two reactions to a competitor's post about carrier invoice audits and a comment on a keyword post about the same problem. Halvorsen Analytics had two, both on your own company page's post about month-end close. Brightline Health had two reactions to company-page posts about payer contracts. Northwind's three signals cluster on one topic, so that account is probably the one to look at first.
The questions that work best are about patterns you'd otherwise piece together by hand:
- Repeat accounts. "Which companies appeared more than once this week, and what did each signal involve?"
- Competitor activity. "What topics keep coming up in this month's signals from competitor pages?"
- One account. "Show me the recent signals from Northwind Freight and summarize what they engaged with."
- Source performance. "Rank our sources by signals in the last 30 days, and flag any active source that finds posts but few signals."
- Topics. "Which topics keep appearing in the signals with the strongest ICP fit?"
- ICP tuning. "Across this month's signals, which ICP dimension most often scored lowest?"
- Account research. "Read the Account Intelligence report for Northwind Freight and list what we still don't know before the call."
- Meeting prep. "Is there anything in this week's signals I should look at before our pipeline meeting?"
It can help with a first message too, working from the signal's suggested opener and talking points.
What the assistant can see
The assistant works from the workspace you choose when you connect. Each signal reaches it with the person's company and LinkedIn profile link, a preview of the post, why the signal matters and the talking points. It also gets the ICP reasoning with a score for each dimension, and the 1 to 10 score the portal shows as Urgency (called strength in the data the assistant reads).
Beyond signals, it can read your sources with their posts found, posts qualified and signals over the last 7 or 30 days, the posts currently on your watchlist, your ICP settings, your Account Intelligence and Person Intelligence reports, the people you monitor, your monthly usage and your destinations. Destination credentials are never shared with it.
What it can and can't do
Everything the assistant does goes through SignalRaven. It never acts on LinkedIn or touches your LinkedIn account, and it can read your sources and ICP without being able to change them.
A few tools go beyond reading, and their descriptions tell the assistant so:
- Intelligence reports. Running an Account Intelligence or Person Intelligence report spends credits when no report exists yet, or when you ask for a refresh of one that's at least 7 days old. Reading an existing report is free.
- Monitoring people. Monitoring someone spends credits on the activity each weekly check reads and on any signal it delivers, until the monitoring period ends or you stop it.
- Destinations. The assistant can create a destination or change an existing one. Either action changes where your signals are sent.
Connecting grants reading and writing together. If you only want answers, decline any tool call that spends credits or changes a destination.
How to connect
SignalRaven's MCP server lives at https://api.signalraven.ai/mcp. Every assistant connects the same way: add that URL as a connector, sign in to SignalRaven, choose the workspace and allow access.
- Claude: open Settings, then Connectors, add a custom connector, paste the URL and connect.
- ChatGPT: the ChatGPT guide walks through both routes, from the directory listing or by adding the server yourself.
- Other assistants and coding tools: the MCP page has the configuration for each one.
A trial or paid workspace gives the assistant your live signals. A workspace that isn't active sees sample data.
When to use the REST API instead
An assistant is good for questions you think of in the moment. A scheduled job is better for anything that has to run the same way every day.
The REST API serves the same data as the assistant, plus the same report, monitoring and destination actions. Use it to:
- Sync signals into a warehouse. Page through new signals on a schedule and join them to your CRM and product data.
- Build an internal dashboard. Pull source performance and usage into the reporting your team already reads.
To send each new signal somewhere as it arrives, use a webhook destination instead. The integrations guide covers the setup.
Workspace admins create keys under Workspace, then API keys. A key carries a client ID and a secret that's shown once, and the same key works for the API and for MCP. The API reference lists every endpoint.
FAQ
Which AI assistants work with SignalRaven?
ChatGPT and Claude, plus other assistants and coding tools that support remote MCP servers. The MCP page lists the setup for each.
Does the assistant use my LinkedIn account?
No. The assistant reads data from your SignalRaven workspace. Nothing it does posts, messages, connects or reacts on LinkedIn.
What can't the assistant see?
Comment text, tags and the specific source a signal came from. It can tell a competitor-page signal from a keyword signal, but not whose page it was.
Can the assistant change my sources or ICP?
No. It can read them. The actions it can take are running Intelligence reports, starting or stopping person monitoring and managing destinations.
Do I need a paid plan?
No. A trial or paid workspace gives the assistant your live signals.