All your AIs, MCPs and skills
— one voice away.
Conduit unifies Claude, every MCP, every skill, your whole memory — into one brain you can talk to from any device. Solo, or as a single shared brain for your whole company.
Free OSS · Cloud 9 €/mo (Q3) · Teams: 4-week paid setup
Connected
Every tool you already use — wired into one brain.
Conduit speaks the MCP protocol. Anything that exposes one — Slack, HubSpot, GitHub, DATEV, Notion, Supabase, Linear, your own internal tools — becomes a verb your Claude stack can do, by voice, from your phone.
Conduit for Teams
Same brain. Same context. Same answer.
One Claude instance for your whole company. Every employee — founder, sales, ops, the new hire who started this morning — talks to the same brain, with the same memory, the same skills, the same wired tools. When the answer changes, it changes for everyone, at the same time.
No more “ask Tobias, he knows”. The knowledge sits in Conduit. Onboarding a new hire takes a prompt, not a quarter.
We build the MCP layer to your DATEV, HubSpot, Lexware, M365. Every employee gets the same access, the same way, by voice or by chat.
When a process changes, the brain learns it once. Nobody is working off last quarter’s playbook by accident.
How a pilot works.
- 1We map your stack and define the first 3 MCPs to wire (typical: DATEV, HubSpot, M365).
- 24-week paid setup: brain, MCPs, skills, memory, per-team scopes. Hosted in Germany or your VPC.
- 3Monthly retainer covers updates + new MCPs. Your Anthropic key, pass-through tokens — no reseller margin.
Tell us about your team.
How it works
Three pieces. No reseller in the middle.
Bridge on your Mac
A tiny daemon spawns your local claude CLI per request. SQLite log, JWT auth, warm-pool process pool. Runs on launchd, restarts itself.
Cloudflare Tunnel
Outbound-only WSS — no open ports, no inbound firewall holes. Your DNS, your cert, your domain. We never see plaintext.
Conduit clients
PWA, native macOS, iOS/Android wrappers, Alexa skill. Live re-attach over WS so a 4G drop never breaks the answer mid-stream.
Why Conduit
What no other tool gives you.
| Conduit | Claude.ai / ChatGPT | Cursor | |
|---|---|---|---|
| Voice in | ✓ | Partial | — |
| Mobile / Alexa / CarPlay | ✓ | Partial | — |
| Access your repos / shell / files | ✓ | — | Editor only |
| Your own MCP servers | ✓ | — | — |
| Your own skills + memory | ✓ | — | — |
| Same brain for the whole team | ✓ | — | — |
| Bring-your-own subscription | ✓ | Anthropic-tied | Reseller markup |
| Data stays on your machine | ✓ | — | — |
Pricing
Three ways in.
OSS is free, forever. Cloud is a thin convenience layer. Teams is hands-on integration.
- Full source on GitHub
- Cloudflare Tunnel + Vercel + Resend (free tiers)
- ~30 min one-time setup
- Bring your own MCPs & skills
- One-click installer · no DNS / Cloudflare / Vercel
- Managed tunnel + auto-updates
- Hosted Whisper + Alexa provisioning
- Multi-bridge auto-discovery
- Priority email support
- One brain for the whole company
- Custom MCPs (DATEV, HubSpot, M365, …)
- Per-team scopes, audit log
- Hosted in Germany or your VPC
- Your Anthropic key, pass-through tokens
FAQ.
Where do the LLM tokens come from?+
Your own Anthropic subscription — the bridge runs the claude CLI on your machine, with your key. Conduit never sees a token of yours and we don't resell Anthropic. Cloud is a SaaS layer for the tunnel and ops, not an LLM proxy.
How can my whole team share one brain?+
Conduit for Teams runs one Claude instance with shared memory, shared skills, shared MCP wiring. Each employee gets a Conduit client (PWA, mobile, Alexa) that authenticates per-user against that brain. Per-team scopes and audit log included.
Does my data leave my machine?+
Chat history is stored in local SQLite on your bridge. Solo Cloud never sees the messages — only the encrypted WebSocket between your phone and your Mac, terminated at Cloudflare. Teams is hosted in Germany (or your VPC) with the same model: tokens pass through Anthropic, knowledge stays with you.
What if Anthropic builds the same thing?+
Likely they will eventually, but it'll be Anthropic-only. Conduit is provider-agnostic — the same architecture works for any local-CLI LLM. We optimise for the day Anthropic ships parity, not against it.
How fast is it really?+
v3.0 keeps a warm claude process per session, so follow-up answers start streaming in ~1.5 s. The WebSocket re-attaches after mobile drops so a 4G blip never breaks a reply mid-flow.