Run your LinkedIn outreach through Claude . You send from your own account, with your approval.
Trained on messages you already sent, Cadenza drafts in your phrasing and keeps your voice across chats, right inside the open-source server you self-host.
Built on linkedin-mcp-server
The open-source MCP server that brings LinkedIn into Claude. Cadenza is the optional hosted layer on top, and every message still goes out from your own account, only after you approve it.
You know exactly who to message. The note is still in your drafts.
There is a list in your head of people worth reaching: the founder who would make a great partner, the candidate who would say yes to the right note, the prospect one good message away from a reply. You open the chat to write the first line, and it goes nowhere.
So you ask an AI to draft it, and it hands you a press release with your name on top. You rewrite the whole thing, which is slower than writing it yourself. Generic openers are the ones people scroll past, and you can feel it before you hit send.
Then the week gets busy, and outreach is the first thing you drop. The warm intro cools, the candidate signs somewhere else, and the thread you meant to restart is three months old.
Your outreach, handled inside Claude.
It runs on the open-source server you already run, with no new tool to learn and no new place for the work to live. Every message still sounds like you.
No new tool to adopt
Drafting and review run as tool calls inside the Claude you already use. There is no separate app to learn.
Free base, premium tools
The open-source server is free in the Claude you already use. Cadenza unlocks the premium tools right there, no setup.
Sounds like you.
Your Voice Profile learns from messages you have already sent. New drafts come out in your phrasing, down to how you open and sign off.
From draft to sent
You approve every draft, then Cadenza sends it through your own connected account.
Stay close to the people who matter
The contacts who move your work forward deserve more than a template. Cadenza drafts in your voice, and you stay the sender.
What you actually get
A prompt forgets when the chat ends. A busy week drops the follow-up. These four hold: your trained voice, first notes beyond your connections, one-approval sends, and a morning brief on who replied.
Get CadenzaCadenza trains it on messages you already sent, then scores drafts against sent messages it held back. It learns habits you would never put in a prompt, like sending several short bubbles instead of one block. It holds across every session and chat.
The free server stays free. Here is what the paid layer adds.
Cadenza is an optional hosted layer for the open-source server you already run. All sixteen tools of that server stay free and maintained under Apache 2.0. You self-host it and keep your own session, with every line open to read. Nothing ever moves from free to paid.
So why pay? You can rebuild the drafting yourself with the free read tools and your own prompts. What a prompt cannot give you is a Voice Profile that persists. The free way means re-explaining how you write in every new chat, and it still reads like a press release.
The paid layer covers the parts you cannot stand up locally: a Voice Profile that lives between sessions, Sales Navigator search and lead lists, a first message to people you are not yet connected to, a staged send through your own connected account after you approve each draft, and reply tracking that tells you who answered. Your subscription also funds maintenance of the free server.
Connect Cadenza
Cadenza runs as a hosted remote MCP server. Connect it once in ChatGPT or Claude on the web, and it works from every device you sign in on. Nothing to install. Nothing requires your desktop.
- 1 In Claude on the web, open Settings, then Connectors, and choose Add custom connector. (opens in a new tab)
- 2 Paste the Cadenza URL and click Add.
- 3 Click Connect, then sign in and subscribe.
- 4 Open a chat and enable Cadenza from the + menu.
Set it up in Claude on the web or desktop, and it works in the iOS and Android apps too.
- 1 In ChatGPT on the web, open the New App form. (opens in a new tab)
- 2 Give it a name and icon, and paste the Cadenza URL as the server URL. Download the Cadenza icon
- 3 Tick the confirmation notice and click Create.
- 4 Sign in and subscribe.
- 5 In a chat, open the + menu and pick Cadenza.
Set it up in ChatGPT on the web. Sending works on web and desktop for now.
The professional layer for the open-source server you already run
Cadenza Premium is the hosted layer: the premium tools and a managed connection, with nothing for you to set up.
The open-source LinkedIn MCP server
- Open-source LinkedIn MCP server
- Self-hosted, connect your own LinkedIn
- Profiles, companies, jobs, and messages
- Community support
Premium outreach tools in your AI client
- Everything in Free
- Draft and send messages in your own voice
- Voice Profile trained on your writing
- Hosted premium tools, no setup
- One connected LinkedIn account
- InMail to people you are not yet connected to
- Sales Navigator search and lead lists
- Recruiter search, talent pools, and pipelines
A shared model trained on your team's voice, so every draft matches your brand.
- Everything in Premium, for every seat on the team
- One shared model trained on your team's past messages and your outreach framework
- A house style your admin sets once, the shared default every draft starts from
- A short setup session to train the first model, then retrain and roll updates to every seat
- A named contact for your team and a private model kept to your organization
Compare plans
| Free €0 | Cadenza Premium €29 / month | Cadenza Enterprise €99 / seat / month | |
|---|---|---|---|
| The open-source base | |||
| MCP read tools (profiles, companies, jobs, messages) | Included | Included | Included |
| Connect your own LinkedIn account | Included | Included | Included |
| Draft with your own prompts | Included | Included | Included |
| Voice and drafting | |||
| Draft and send messages | Not included | In your voice | In your team's voice |
| Voice Profile | Not included | Personal | Shared team model |
| Common ground openers | Not included | Included | Included |
| Delivery | |||
| Reliable delivery | Not included | Included | Included |
| Reply status you can check | Not included | Included | Included |
| Follow-up drafts | Not included | Included | Included |
| InMail and Sales Navigator | Not included | Included | Included |
| Team and support | |||
| Connected accounts | Your own | 1 | 3+ seats |
| Shared house style your admin sets | Not included | Not included | Included |
| Team model retraining and rollout | Not included | Not included | Included |
| Private model kept to your organization | Not included | Not included | Included |
| Support | Community | Standard | Named contact |
| View on GitHub | Set up your team model | ||
Questions worth answering before you connect an account
Still have questions? Talk to us
Cadenza is an independent product. It is not affiliated with, endorsed by, or sponsored by LinkedIn Corporation. LinkedIn is a trademark of LinkedIn Corporation.
Start writing in your voice
Connect your account and let Cadenza learn your style from your past messages. Then send notes that actually sound like you.
- Handled from draft to sent
- Runs inside the Claude you already use
- Goes out through your own account