The connector says it needs authentication or returns 401.

Your session expired or the key was revoked. In chat clients, remove and re-add the connector to run the sign-in again. In Claude Code, run /mcp and re-authenticate. For API keys, create a fresh key under Settings, then Connections.

I get a 403 saying MCP access is not enabled.

MCP is included with every ThreadScope plan, so a 403 on the base tools means something is off with your account rather than your plan - contact support and we will sort it out.

My results are capped at 10 posts, or the since filter is refused.

Larger result sets and time filtering are part of the Patrol plan and above. The free taste is deliberately useful but small - upgrade on the pricing page to lift the cap.

My client is stuck in an OAuth loop.

Clients that use the mcp-remote bridge cache credentials in ~/.mcp-auth. Delete that directory and reconnect. Also check your machine clock - OAuth is time-sensitive.

Can the connector post to Reddit or change my data?

No. Every tool is read-only. ThreadScope never posts on your behalf; you always click through to Reddit and reply as yourself.

Docs · MCP server

Watch Reddit from Claude.

ThreadScope's MCP server connects your AI - Claude, Cursor, Codex, ChatGPT or anything that speaks MCP - to your triage inbox. Ask what's new, what's worth replying to, and what ThreadScope drafted. Read-only, always: your AI can look, never touch.

Server URL

https://threadscope.io/api/mcp

Chat clients sign in with OAuth. Config-file clients use an API key. No account yet? Connecting walks you through creating one - or start free now.

Setup · Two minutes

Connect your client

claude.ai and Claude Desktop

Open Settings → Connectors and choose Add custom connector.
Paste the server URL: https://threadscope.io/api/mcp
Click Add, then Connect. Your browser opens ThreadScope - sign in (or create an account) and approve.

The claude.ai custom connector dialog with the ThreadScope URL pasted in.

screenshot pending - /docs/mcp/claude-connector-dialog.png

ThreadScope's authorisation screen showing what Claude will be able to read, with Deny and Approve buttons — You approve exactly what Claude can read. Revoke any time from Settings.

On the Claude free plan you can add one custom connector; paid plans allow more. Claude Desktop uses the same account-level connector - add it once, use it in both.

Claude Code

Add the server, then authenticate in your session:

claude mcp add --transport http threadscope https://threadscope.io/api/mcp

Run /mcp inside Claude Code to complete the OAuth sign-in. For headless or CI use, create an API key under Settings → Connections and pass it as a bearer token instead:

claude mcp add --transport http threadscope https://threadscope.io/api/mcp \
  --header "Authorization: Bearer ts_your_key_here"

Cursor

Add to Cursor with one click - or add it by hand in ~/.cursor/mcp.json:

~/.cursor/mcp.json

{
  "mcpServers": {
    "threadscope": {
      "url": "https://threadscope.io/api/mcp"
    }
  }
}

Cursor prompts for the OAuth sign-in on first use.

VS Code (Copilot)

Install in VS Code - or add it to .vscode/mcp.json (note VS Code uses servers, not mcpServers):

.vscode/mcp.json

{
  "servers": {
    "threadscope": {
      "type": "http",
      "url": "https://threadscope.io/api/mcp"
    }
  }
}

Codex

codex mcp add threadscope --url https://threadscope.io/api/mcp

Codex uses OAuth by default (codex mcp login threadscope). For an API key instead, configure ~/.codex/config.toml:

~/.codex/config.toml

[mcp_servers.threadscope]
url = "https://threadscope.io/api/mcp"
bearer_token_env_var = "THREADSCOPE_API_KEY"

ChatGPT

ChatGPT (web, Plus/Pro and above): enable Settings → Apps → Advanced settings → Developer mode.
Add a connector with the server URL https://threadscope.io/api/mcp and choose OAuth.
Approve access on the ThreadScope screen that opens.

ChatGPT's developer-mode connector form with the ThreadScope URL.

screenshot pending - /docs/mcp/chatgpt-connector.png

Everything else

Any MCP client that supports remote servers works: point it at https://threadscope.io/api/mcp (streamable HTTP) and sign in with OAuth. If your client can't do OAuth, send an API key as Authorization: Bearer ts_... - create and revoke keys under Settings → Connections.

ThreadScope's Connections settings showing the MCP server URL, an API key, and a connected app with revoke buttons — Settings → Connections: your server URL, API keys, and connected apps.

Capabilities

What your AI can do

Five tools, all read-only. Your AI sees ThreadScope's triage - matches, tags, statuses, drafts - and links back to Reddit for the threads themselves.

Tool	Access	What it does
list_matches	Read	Your triaged inbox: matched posts with tags, status and permalinks. Filter by status, tag, subreddit or time.
get_match	Read	One match in full, including the drafted reply ThreadScope wrote (Patrol plan and above).
get_counts	Read	Inbox counts by status - the quick opener for a daily brief.
list_tags	Read	Your tag definitions and search terms, so your AI can explain why something matched.
list_subreddits	Read	The subreddits you're monitoring.

Free and Scout plans get a capped taste; Patrol and above unlock full result sets, time filtering and drafted replies. See plans.

Try it

Prompts that earn their keep

“What's new in my ThreadScope inbox? Give me the five most worth replying to.”
“Anything urgent in my monitored subreddits since Monday?”
“Which of my tags matched the most this week?”
“Show me the drafted reply for that top match, then open the thread.”
“Every morning at 9, check my ThreadScope matches and brief me.”

Data & security

Read-only by design. No tool can change your data or post anywhere. Replying always happens on Reddit, as you.
Your triage, not a content feed. The connector returns ThreadScope's own work - match metadata, tags, statuses and drafts - with permalinks so you read threads on Reddit itself.
Revocable in one click. Connected apps and API keys live under Settings → Connections; revoking kills their access immediately.
Treat thread content as untrusted. Reddit text flowing into any AI can contain prompt injection. ThreadScope screens matches for injection patterns at ingestion, and keeping your client's tool-use confirmations on is good hygiene with every MCP server, ours included.

Troubleshooting

When it doesn't connect

The connector says it needs authentication or returns 401.: Your session expired or the key was revoked. In chat clients, remove and re-add the connector to run the sign-in again. In Claude Code, run /mcp and re-authenticate. For API keys, create a fresh key under Settings, then Connections.
I get a 403 saying MCP access is not enabled.: MCP is included with every ThreadScope plan, so a 403 on the base tools means something is off with your account rather than your plan - contact support and we will sort it out.
My results are capped at 10 posts, or the since filter is refused.: Larger result sets and time filtering are part of the Patrol plan and above. The free taste is deliberately useful but small - upgrade on the pricing page to lift the cap.
My client is stuck in an OAuth loop.: Clients that use the mcp-remote bridge cache credentials in ~/.mcp-auth. Delete that directory and reconnect. Also check your machine clock - OAuth is time-sensitive.
Can the connector post to Reddit or change my data?: No. Every tool is read-only. ThreadScope never posts on your behalf; you always click through to Reddit and reply as yourself.

Delegate the watching.

ThreadScope triages your subreddits around the clock. Your AI turns it into a brief. You just reply well.

Start free →