Major x Snitchfeed: Official API and MCP Connectors for Social Listeni

Snitchfeed is a social listening tool that scores public posts for intent, and its official API and MCP connectors are now live on Major. Attach the connector, and an agent can set up the listeners while an app polls mentions, stores the ones above your fit-score cutoff, and notifies an owner. Major is where that loop becomes governed software instead of a prompt that re-runs every morning.

Social listening usually dies in a Slack channel. Mentions pile up, nobody scores them, and the useful ones get buried under brand spam. Snitchfeed scans LinkedIn, X, Reddit, Bluesky, and Hacker News, and returns each match with intent tags, a fit score, sentiment, and a link to the source post. This article is for RevOps, growth, and content owners who want those signals to land in a system of record, not a channel.

What the Snitchfeed connectors give you

There are two ways in, and they suit different jobs.

The HTTP API sits at api.snitchfeed.com/v1 behind a Bearer token. Reading your org's mentions, listeners, feeds, and usage does not spend search credits. Metered search covers X, LinkedIn posts and people, LinkedIn company pages, and Hacker News, with a limit of 20 requests per minute per search surface and at most two in-flight requests across v1.

The MCP server sits at api.snitchfeed.com/mcp and authenticates with OAuth, so there is no key to paste. It exposes 42 tools: read mentions, manage listeners, build saved feeds, run ad hoc search (including LinkedIn person posts, comments, and reactions), and pull analytics.

Use MCP when an agent should operate the workspace in language: create a listener, tighten a noisy query, ask for unseen high-fit mentions from the last day. Use the HTTP API when an app should poll the same data on a cron, write rows to a database, and fan out updates without another reasoning pass.

Build the Snitchfeed mention queue on Major

The pattern is the same for every workflow below. The agent reasons once about the setup, then builds an app that runs it.

1. Connectors. Attach the Snitchfeed API connector for scheduled polling and the Snitchfeed MCP connector for the agent. Add Slack for notifications and, depending on the workflow, a CMS, Taplio, or a CRM.
2. The app that holds state. A managed database table stores mention id, platform, author, URL, intent tags, fit score, raw text, and a status of new, assigned, or dismissed. Storing the id means the same post never alerts twice.
3. The agent's judgment. Which queries collide with noise, where the fit-score threshold sits, which new account belongs on a target list, and whether a spike is a real shift or one loud thread. That is the part that needs a model.
4. The deterministic loop. Poll, filter above threshold, dedupe, write, notify. This runs the same way every time, so it lives in code. It does not go back through the model on each run.
5. The trigger. A scheduled workflow runs the app every morning, or on whatever cadence the listener volume justifies.
6. Governance. Credentials are scoped to the connector, access to the queue is role-based, and every write to the app is audit-logged. IT can see what the agent changed and who acted on a mention.

The result is front-loaded then flat. The agent spends tokens designing the loop once, and the daily polling no longer costs model time. Start building free and ask an agent to stand up the queue.

Five workflows to build first

Snitchfeed keywords to a published page

Keywords from public conversation land in an SEO app on Major, then a chain of agents takes them live. A research agent scores them for volume and keeps the ones worth a page. A brief-builder agent writes the brief in the same app. A writer agent drafts. A publisher agent, connected to the CMS, takes the draft live.

The app holds the keyword, brief, draft, and live URL, so the same theme never gets written twice. Answer engines cite pages that match how people actually ask, which is why the keywords start in public posts and not a planning spreadsheet. The failure mode is volume bias: a loud thread produces a keyword that nobody searches for, so the research agent needs a volume check before anything reaches a brief.

Trend monitoring and keyword extraction

A scheduled agent asks Snitchfeed for mention volume by platform, the top intent tags, and the phrases that keep recurring, then writes those into a workbench a writer can open tomorrow. We already run this internally: Snitchfeed listeners feed specific themes into daily SEO keyword picks. The useful output is a growing list of phrases buyers used this week, each linked to the post.

LinkedIn posts from live signals

A scheduled agent pulls high-fit mentions and clusters them into topics a founder or marketer could post about. A content agent drafts each post in the house voice, and Taplio schedules it. Snitchfeed does not write the post, the content agent does not scrape LinkedIn, and Taplio does not invent the topic. A person can still kill a draft before it goes out.

Reddit replies for answer engine visibility

Reddit threads are where people ask the follow-up questions answer engines later cite. Snitchfeed listens to those threads, and an agent writes a reply draft against each topic. The app stores the source thread, the suggested answer, and a status of ready or skipped.

A person publishes the reply. Auto-posting to Reddit violates its anti-botting policy and can get the account banned, so the app is a queue of drafts and never a poster.

ABM account listening on LinkedIn and X

Named-account listening is a different query than brand monitoring. Point listeners at the companies on your target list and at the people in the buying committee. Snitchfeed can search LinkedIn posts, pull posts and comments from a profile URL, scrape comments and reactions on a given post, and search X for the surrounding conversation.

The agent decides which accounts and people belong on the list and which intent tags count as sales-ready: asking for a tool, comparing vendors, complaining about a competitor. The app polls, scores, dedupes, and drops high-fit rows into a prospecting queue or CRM. A human still takes the meeting. They should not be the one searching LinkedIn at 7am.

Where to start

Start with one listener you care about: a category question, a competitor name that does not collide with common words, or a short list of target accounts. If you do not have a Snitchfeed workspace yet, sign up for Snitchfeed first, since paid tiers include the API and MCP. Then attach the connector on Major. If you already have Snitchfeed, attach it today.

Start building on Major

Attach the Snitchfeed connector, let an agent build the mention queue once, and run it on a schedule with scoped credentials and an audit trail on every write.

• Start building free
• Book a demo

Related articles

• Anthropic vs OpenAI: An API Comparison for Builders
• Asana Automation: From Rules to an Agent That Reads Across Projects
• Jira Automation for Engineering-Ops AI Agents