The ParseStream MCP serverfinds buying-intent conversations and drafts your reply

1. 01get_brand_context

   “What ParseStream brand am I connected to, and which keywords are you tracking?”

   The opening move of every session, and the fastest way to confirm the connection works. You get the brand, its competitors, every keyword with the platforms it runs on, and the credit balance — so nothing later in the chat is guessed.

2. 02search_mentions

   “Find mentions from the last 7 days that I haven't handled yet, relevance 4 or higher, where someone sounds like they're actively looking for a tool.”

   That last clause is the whole product. The assistant turns it into intent filters — BUY_INTENT and PRODUCT_QUESTION — against a 7-day window, confident keyword matches only, and the untouched bucket. What comes back is a shortlist, not a firehose.

3. 03get_mention

   “Show me the second one in full — the whole comment, the upvotes, and whether there's already a draft for it.”

   Search gives you enough to triage. Upvotes, subreddit size, the engagement-opportunity flag and whether a draft already exists only come back here, on the one mention you asked for.

4. 04draft_reply

   “Draft a short, value-first reply to it.”

   The only step that costs anything: 5 credits, five to twenty seconds, and the text comes back in the chat. It is also saved to your Drafts queue. Nothing is posted anywhere.

5. 05Your turn — no tool

   (you read it, edit it, and post it from your own account)

   This step is not automatable and is not meant to be. Open Replies in ParseStream, make it sound like you, and post it yourself on the platform it belongs to. No tool on this server can do it for you.

6. 06mark_mention_replied

   “Mark that one replied, and star the other three so I can come back to them tomorrow.”

   The flags are the same ones the web app shows, so the queue you see tomorrow already reflects this conversation. Starring uses highlight_mention — and both take a mention out of the untouched bucket.

Six turns, six of the seven tools, one credit charge. Everything except the draft is free, and the only thing that ever reaches Reddit is the comment you post yourself from your Replies queue in ParseStream.

get_brand_contextFreeread-only

The whole workspace: brand profile and competitors, every tracked keyword with its ID and the platforms it is monitored on (paused ones included, flagged rather than hidden), the current credit balance and whether drafting is available at all.

search_mentionsFreeread-only

Searches your mentions across all five platforms. Filter by timeframe, platform, intent, relevance score, keyword, subreddit, mention type or status. Up to 30 per page, newest-posted first.

get_mentionFreeread-only

One mention in full: the untruncated body, the parent post title, the author, upvotes, subreddit size, the relevance score and intent tag, and whether a draft is already waiting for it.

draft_reply5 creditswrites a draft

Generates a reply draft for one mention and saves it to your Drafts queue. Pick the approach and the length, or let it use the brand's saved defaults. It returns text — a human still opens Replies, edits it and sends it.

mark_mention_repliedFreewrites a flag

Sets or clears the replied flag on one mention — the same toggle the web app shows. It moves the mention out of your new queue and leaves any existing draft alone.

highlight_mentionFreewrites a flag

Stars or unstars one mention, so a shortlist survives past the end of the chat. Starring also takes a mention out of the new queue — that filter means untouched, not unread.

get_brand_detailsFreeread-only

Brand name, description and website, and nothing else. One indexed row, back in milliseconds — the tool an assistant reaches for when all it needs to know is who it is working for.

Why is get_brand_details last? Because the tool list is served in a pinned order so the response is byte-identical between requests and clients can cache it. New tools get appended, never inserted — inserting one would invalidate every client’s cache from that entry on. Only draft_reply is billed, at 5 credits, and it needs an active subscription or trial; everything else works on any plan.

claude mcp add --transport http parsestream https://mcp.parsestream.com/mcp \
  --header "Authorization: Bearer ps_live_xxxxxxxxxxxxxxxx" \
  -s local

{
  "mcpServers": {
    "parsestream": {
      "url": "https://mcp.parsestream.com/mcp",
      "headers": {
        "Authorization": "Bearer ps_live_xxxxxxxxxxxxxxxx"
      }
    }
  }
}

{
  "servers": {
    "parsestream": {
      "type": "http",
      "url": "https://mcp.parsestream.com/mcp",
      "headers": {
        "Authorization": "Bearer ps_live_xxxxxxxxxxxxxxxx"
      }
    }
  }
}

[mcp_servers.parsestream]
url = "https://mcp.parsestream.com/mcp"
bearer_token_env_var = "PARSESTREAM_API_KEY"

Connected? Ask your assistant “what ParseStream brand am I connected to?” — it should name your brand and list your tracked keywords. If it does not, the header is nearly always the cause; the full documentation has the troubleshooting list.