MCP Remote MCP server and REST API

Every podcast and YouTube transcript, cited to the second.

Plug CleoLM into Claude, ChatGPT, Grok, Cursor or your own code. Your model does the reasoning; CleoLM supplies the source: searchable, timestamped transcripts with a deep link to the exact moment each line was said.

Connect your agent

MCP https://cleolm.com/mcp

No account yet? Try it now, no sign-up:

TRY https://cleolm.com/mcp/open

Search and passages across all 158 channels, limited to 30 calls a day. Connect /mcp for transcripts, ingest and your 50 free monthly credits.

The Peter Attia Drive · #47, Matt Walker on Sleepget_passage

56:09constitutively sleeping six hours a night would suffer perhaps fewer
56:14of the memory consolidation issues because they're still getting much of their deeper
56:20sleep earlier in the evening presumably where that hippocampal consolidation takes place
56:42is that you need sleep after learning to essentially hit the Save button on new memories
56:47so that you don't forget and for that it's all about deep non REM sleep

youtube.com/watch?v=VkX3nYpMr2g&t=3402s

158 channels, continuously synced ~4,000 new episodes a month ~3.1M indexed passages Any video or episode, on demand

01 / Install

One endpoint. Every agent you already use.

Pick your client. Most sign in with OAuth; headless setups use a scoped API key from /developers.

claude.ai · Desktop · Cowork

Add a custom connector Directory listing soon

Claude has no install link for connectors yet, so paste the URL once. It syncs to Claude on mobile too.

Settings → Connectors → Add custom connector.
Name CleoLM, URL https://cleolm.com/mcp.
Click Connect, sign in to CleoLM, then Allow.

Terminal

One command, then sign in

Run this, then type /mcp in Claude Code, pick cleolm and choose Authenticate.

claude mcp add --transport http cleolm https://cleolm.com/mcp

# just trying it? no account, no sign-in

claude mcp add --transport http cleolm-open https://cleolm.com/mcp/open

# headless or CI: use an API key

claude mcp add --transport http cleolm https://cleolm.com/mcp \
  --header "Authorization: Bearer clm_live_…"

chatgpt.com

Add it in developer mode Directory listing soon

Until the app directory listing is live, ChatGPT connects custom MCP servers through developer mode.

Settings → Apps & Connectors → Advanced → turn on Developer mode.
Create: name CleoLM, URL https://cleolm.com/mcp, authentication OAuth.
Sign in to CleoLM and click Allow.

Custom GPTs: Actions → Import from URL → https://cleolm.com/api/v2/openapi.json, auth Bearer clm_live_….

Codex CLI · IDE extension

Add, then log in

Two commands. The second opens your browser to sign in to CleoLM.

codex mcp add cleolm --url https://cleolm.com/mcp
codex mcp login cleolm

# or ~/.codex/config.toml (CI: export CLEOLM_API_KEY)

[mcp_servers.cleolm]
url = "https://cleolm.com/mcp"
bearer_token_env_var = "CLEOLM_API_KEY"

Cursor

One-click install

Opens Cursor with CleoLM pre-filled. Sign in when Cursor prompts.

Add to Cursor

// or ~/.cursor/mcp.json

{
  "mcpServers": {
    "cleolm": { "url": "https://cleolm.com/mcp" }
  }
}

With an API key instead, add "headers": { "Authorization": "Bearer ${env:CLEOLM_API_KEY}" }. Never commit a literal key.

VS Code · Copilot agent mode

One-click install

Opens VS Code's MCP install prompt for CleoLM.

Install in VS Code

// or .vscode/mcp.json

{
  "servers": {
    "cleolm": { "type": "http", "url": "https://cleolm.com/mcp" }
  }
}

Windsurf: "serverUrl" in ~/.codeium/windsurf/mcp_config.json. Gemini CLI: "httpUrl" in ~/.gemini/settings.json.

grok.com · Grok Bot · xAI API

Add a custom connector

grok.com → Connectors → New → Custom, paste the URL and choose OAuth. Business admins can add it for the whole team.

Grok Bot: add an MCP server named cleolm with URL https://cleolm.com/mcp and header Authorization: Bearer clm_live_….

# xAI Responses API, remote MCP tool

"tools": [{
  "type": "mcp",
  "server_url": "https://cleolm.com/mcp",
  "server_label": "cleolm",
  "allowed_tools": ["search", "get_passage"],
  "authorization": "clm_live_…"
}]

Meta Muse

Ask Muse to connect

Muse has no connector catalog yet. Paste this into a Muse chat and sign in to CleoLM when asked.

Connect to the MCP server at https://cleolm.com/mcp

LM Studio

One-click install

Adds CleoLM to LM Studio's MCP servers. Your local model calls the tools.

Add to LM Studio

// or Program → Install → Edit mcp.json

{
  "mcpServers": {
    "cleolm": { "url": "https://cleolm.com/mcp" }
  }
}

Goose

One-click install

Opens Goose with CleoLM as a remote extension.

Add to Goose

Run goose configure → Add Extension → Remote Extension.
Name cleolm, URL https://cleolm.com/mcp.

REST · n8n · Zapier · Make

Plain HTTPS with a bearer key

The same tools at https://cleolm.com/api/v2. OpenAPI-aware tools can import the spec directly.

Create an API key OpenAPI spec

curl https://cleolm.com/api/v2/search \
  -H "Authorization: Bearer clm_live_…" \
  -H "Content-Type: application/json" \
  -d '{"query": "hiring your first engineer", "limit": 5}'

02 / How it works

Your model reasons. CleoLM brings the receipts.

There is no LLM on our side. Your agent decides what to look up, CleoLM returns the exact passages, and every claim in the answer links back to the second it was said.

01You askin plain language, in the client you already use.
02Your agent calls searchacross the indexed channels and your library.
03CleoLM returns passageswith timestamps and deep links.
04Your model answersand cites each point.

agent · cleolm connected

You

What have founders said on podcasts about hiring their first engineer? Cite the moments.

tool callcleolm.search1 credit

{ "query": "hiring your first engineer", "scope": "public", "limit": 3 }

result3 passagesdeep links included

47:12Founders Hour · Ep. 88, Building the first teamHire for the problem you'll have in a year, not the one you have today.youtube.com/watch?v=…&t=2832s
1:03:40Build Notes · Seed to Series AOur first engineer was a generalist who'd already shipped something alone.youtube.com/watch?v=…&t=3820s
18:05The Operator · Hiring before product-market fitDo the job yourself for a month before you hire for it.podcast episode · t=1085s

Assistant

Three patterns come up. Hire for next year's problem, not today's 47:12. Favor a generalist who has shipped something solo 1:03:40. And do the work yourself before you hire for it 18:05.

Illustrative session. Show names and quotes are examples, not real search results.

03 / Tools

Nine tools. Priced per call.

The same nine tools over MCP and REST. Read-only except ingest, which only adds to your library. 1 credit = $0.02.

CleoLM tools and credit costs
Tool	What it does	Credits
`search`	Semantic and keyword search across 158 channels and your library. Returns timestamped passages with deep links.query, scope, channels, published_after, published_before, item_ids, limit	1
`discover`	Find YouTube videos, podcast shows or episodes on the open web, beyond the index.query, type, limit	1
`list_episodes`	Latest uploads of a channel or episodes of a podcast feed, newest first.source, limit	Free
`get_transcript`	Full timestamped transcript of an item, paginated (~30k characters a page).id \| url, cursor, max_chars, format	0.2 / page
`get_passage`	The transcript window around a moment, for context on a search hit.id \| url, at_seconds, before_seconds, after_seconds	Free
`ingest`	Transcribe any YouTube video or podcast episode into your searchable library. Premium high-accuracy ASR on Pro and Max.url, quality, wait_seconds	0–36 / hr
`get_job`	Status of an ingest job: queued, running, completed or failed.job_id	Free
`list_channels`	The channels search covers, optionally filtered by name.query	Free
`account`	Your plan, credits remaining and reset date.	Free
`show_player`	An inline player in the chat with the cited moments as clickable timestamps (Claude and ChatGPT).id \| url, moments	Free

04 / Try it

Prompts to paste once you're connected

Click one to copy it.

05 / Auth & spend

You stay in control of access and spend

OAuth or scoped keys

Claude, ChatGPT and Grok sign in with OAuth 2.1 and PKCE. Headless setups use clm_live_… keys: shown once, scoped to allow or block ingest, and revocable on /developers within a minute.

A hard spend cap, by default

When credits run out, calls stop with credits_exhausted. There is no overage billing, and your agent is told to stop rather than retry.

Your model does the reasoning

CleoLM runs no LLM over your requests. It receives only tool arguments, such as a query or a URL, and never your conversation.

Prepaid only

Credit packs are bought up front and never auto-recharge. You always know the most you can spend.

06 / Pricing

Simple plans, metered by credit

Start free: every account gets 50 credits a month to search, cite and transcribe captioned videos. Pro and Max raise the allowance and unlock podcasts, any video and premium transcription.

Pro

$25 / month

1,000 credits included each month
All 9 tools over MCP and REST
Premium high-accuracy ASR for ingest

Choose Pro

Max 5× the credits

$99 / month

5,000 credits included each month
All 9 tools over MCP and REST
Premium high-accuracy ASR for ingest

Choose Max

Each search costs 1 credit (~2¢).

Credit costs

Credits per operation
search, discover	1
get_transcript, per page	0.2
get_passage, list_episodes, list_channels, get_job, account, show_player	0
ingest, already transcribed	0
ingest, YouTube captions	4
ingest, podcast audio	10 / hr
ingest, ASR	20 / hr
ingest, AI video	36 / hr

Credit packs

Top up when you need more. Packs are prepaid, used after your monthly allowance, and never auto-recharge.

Credit pack prices
1,000 credits	$20
5,000 credits	$90

Ingest reserves the worst case up front and refunds the difference when the job finishes.

07 / For agents

Reading this as an agent?

Machine-readable entry points: a plain-text summary of the endpoint, auth, tools and pricing, plus the full OpenAPI 3.1 spec.

curl https://cleolm.com/llms.txt

08 / FAQ

Questions

Which plan do I need?

None to start. Every CleoLM account gets 50 free credits a month in Claude, ChatGPT or any agent, covering search, transcripts and captioned YouTube videos. For podcasts, videos without captions, premium transcription and a monthly allowance, choose Pro ($25/month, 1,000 credits) or Max ($99/month, 5,000 credits).

What happens when I run out of credits?

Tool calls stop with credits_exhausted and your agent is told to stop. Nothing is billed automatically. Buy a 1,000 or 5,000 credit pack on /developers, or wait for your monthly reset. The account tool shows both.

Does CleoLM see my conversation?

No. CleoLM runs no LLM over your requests. It receives the tool arguments your model sends, such as a search query or a video URL, and returns transcript data. Reasoning and writing happen in your own model.

Which channels are covered?

158 YouTube and podcast channels are synced continuously, about 4,000 new episodes a month. Call list_channels for the current list. Anything else can be added on demand with ingest, which puts it in your private searchable library.

How accurate are the transcripts?

They are speech-to-text, so expect occasional recognition errors, and quote with that in mind. On Pro and Max, ingest with quality: "premium" re-transcribes with high-accuracy ASR. It helps most with multi-speaker interviews and podcasts.

Should I use OAuth or an API key?

Use OAuth wherever the client supports it: Claude, ChatGPT, grok.com, Codex, Cursor and VS Code. Use an API key for Grok Bot, the xAI API, CI jobs, n8n, Zapier and your own code. Keys are shown once and can be revoked at any time.

Is CleoLM in the Claude or ChatGPT directories?

Not yet; the listings are coming soon. A custom connector with https://cleolm.com/mcp works today in both.

Give your agent something worth citing.

Connect in under a minute. Every answer links to the moment it came from.

Connect your agent Get an API key