OAuth for your MCP server, the way claude.ai wants it.
Point us at an MCP server, an API, or a SQL database. We front it with OAuth that claude.ai and ChatGPT accept, run every call as the person who asked, and log every one.
- Works with n8n, Make, FastMCP or your own code
- Runs as the user, not a shared token
- Your server stays where it is, including inside your own network
Free for your own MCP server. No card.
One URL your users install. OAuth on the outside, your own identity provider on the inside, your server untouched.
It works in Claude Desktop and fails on claude.ai. Here is why.
No authorization server.
Desktop accepts a bearer token. claude.ai runs a full OAuth discovery, and most servers, n8n's MCP trigger node included, do not expose one. Nothing you configure fixes that.
The token disappears on redirect.
Cross-host redirects drop the Authorization header. Private IP ranges, split-horizon DNS and AAAA-only hostnames are refused before the request ever reaches your logs, which is why the log is empty.
It connects, then dies.
SSE buffering behind a tunnel. Chunked encoding, proxy trust, Host header. Three settings, and the order matters.
We solved these once. You stop solving them per server. The full picture, with all three failure modes and five approaches compared: MCP authentication for claude.ai →
It carries your name, not ours.
Your people don't install "Loopthink". They install your company's connector, on your own domain, with your logo, sitting in the same list as every other connector in Claude and ChatGPT. Rolled out once, for everyone.
- Your name, your logo. The connector is published on your own domain, so ChatGPT and Claude show your brand, not a vendor nobody recognises.
- One URL for the whole company. No per-system setup for employees. They add the connector once, sign in with SSO, and the tools they are allowed to use appear.
- Tools follow permissions. A user only ever discovers the tools their role grants. Revoke the role, and the tools are gone from their next request.
- Nothing installed locally. No desktop agent, no VPN client, no API keys handed to the AI vendor. Authentication runs over OAuth with short-lived tokens.
Included from €49 per month, on the Team tier.
Acme Group
mcp.acme.com · signed in as j.weber@acme.com
- sap_invoices_query
- crm_contacts_get
- warehouse_stock_scan
+ 11 more, based on your role. Masked fields are marked in every result.
Some other connector
Not governed · not your data
What your employees see. One entry, your brand, their permissions.
Four checks between the model and your data.
MCP made it trivial to plug an AI into a production system. The gateway is what makes it defensible: every single tool call takes the same path, whether it came from ChatGPT, Claude, Langdock or your own UI.
One gateway between every MCP server and every AI client. Who asked, what they were allowed to see, what left your network, and one log of all of it, across Claude and ChatGPT.
- 01
Identify
OIDC/SSO against your own IdP, including Microsoft Entra. The call runs as the person who made it, never as a shared service account.
- 02
Authorize
Per-role, per-tool, read or write. Anything not explicitly granted never shows up in the tool list in the first place.
- 03
Mask
Field-level masking and pseudonymization applied on your side, before the result travels. The model never sees the raw field. Composio, AnythingMCP and both provider tunnels have no field-level masking at all.
- 04
Log
Who asked, which tool, which records, which fields were masked. One audit trail across every AI client, exportable for review.
One policy, every client. The rules live in the gateway, not in the AI tool. Add a client, and it inherits the same access rules, masking and audit trail. Nothing to re-implement per vendor.
Three ways in.
Bring your MCP server
Live in minutesBuilt by n8n, Make, Microsoft or your own team. Point it at Loopthink and it inherits OAuth that Claude and ChatGPT actually accept, one access per person instead of one shared token, and a full audit trail. The server stays where it is, even inside your network.
Free, up to five connectors
Have us generate it
Same dayYou have an OpenAPI spec, a reachable REST API, a SQL database or a Search Console property. We turn it into an MCP server: point us at it, name the tools, mark the personal fields. Live the same day.
From €49 per month
Have us build it
Five working daysNo spec, an unusual protocol, or nobody with the time. We build the connector, name the tools, mask the personal fields and hand it over live. Nothing to pay if we miss the date. Optional on every tier.
€990 once
Systems that are not reachable from outside run behind your own Runner, inside your network. +€199 per month, on any tier.
Internal ERP, nothing exposed: €1,238 to start, €248 per month. That is Team at €49, the private runner at €199, and the first connector built for you at €990.
Internal systems that never touch the internet, and everything else.
One connector in front, two very different worlds behind it. Both governed the same way.
Internal systems
Never exposedSAP, Dynamics, Navision, your databases and internal APIs. The Loopthink Runner sits inside your network and only dials out: it picks up approved calls, executes them locally and returns the result over the same encrypted channel. No inbound port, no VPN, no firewall change. What carries the weight is what happens inside that channel: system credentials never leave your intranet, and personal fields are masked before any result travels.
SaaS & cloud
Same policySalesforce, HubSpot, ServiceNow, Microsoft 365, Jira, GitHub, Stripe and anything else that speaks MCP or REST. Connected directly, but under exactly the same access rules, masking and audit trail as your internal systems, in the same connector.
Already running n8n? Then you already have the runner.
Install one community node and anything your workflows can reach becomes a governed tool: data tables, a Postgres behind the firewall, an internal API. Outbound only, credentials stay in n8n, masking runs in your network.
ERP & Finance
- SAP
- Dynamics 365
- Navision
- DATEV
CRM & Sales
- Salesforce
- HubSpot
- Dynamics CRM
Service & ITSM
- ServiceNow
- Zendesk
- Jira Service Management
Collaboration
- Microsoft 365
- Slack
- Jira
- Confluence
- SharePoint
HR
- Personio
- SuccessFactors
- Workday
Commerce & Payments
- Shopify
- Stripe
- PayPal
Engineering
- GitHub
- GitLab
Databases & Warehouses
- PostgreSQL
- MySQL / MariaDB
- MS SQL Server
- Oracle
- MongoDB
Automation platforms
- n8n
- Make
If it speaks MCP, it already works.
Including the MCP servers your n8n and Make instances generate today. If it only has a REST API, we connect that too, in the same connector, under the same policy.
Don't see your system?
Then we build it. Point us at it. We build the connector, name the tools, mask the personal fields and hand it over live in five working days. €990 once, and nothing if we miss the date.
Live in five working days, or the same day if you connect it yourself.
No platform migration, no rip-and-replace. You start by seeing what is already connected, then put one governed layer in front of systems that already exist.
Kick-off and identity
We connect your identity provider and see what AI already touches in your systems, including the connections nobody reported.
Access
Either you connect your API yourself, or we install the Runner inside your network. Outbound only.
Tools and masking
We name the tools your people will see, set the role matrix and mark every personal field.
Live
Your connector, on your domain, in Claude and ChatGPT. First tool call logged.
Expand
New systems, more teams, through your own approval workflow.
If the first tool call isn't live on day five, the setup fee is on us.
You could build this. Then you'd run it.
One database, one team, a developer who enjoys this: a weekend gets you an MCP server in a container, a tunnel and a token in a config file. That works, and we won't pretend otherwise.
It stops working when the second system arrives, and again when the second assistant does.
Build it yourself
Two systems × three role profiles = six servers
6 containers · 6 tokens · 6 logs · 24 approvals
every server is added and approved in every assistant, one by one
With Loopthink
One connector, four assistants
one identity · one policy · one log
Two systems. Three role profiles, because in an AI client a connector is either on or off and there is no layer underneath it. Four assistants. Two times three is six servers to deploy, patch and log, and six times four is twenty-four registrations to approve, because every server is added to every assistant separately. Five systems on those same three role profiles is fifteen servers and sixty registrations, and one list to work through every time somebody leaves.
| Build it yourself | With Loopthink | |
|---|---|---|
| What you deploy | One server per system, per role profile, per assistant | One connector, on your own domain |
| Who the target system sees | One service account, the same for everyone | The person who asked |
| Who can use which tool | Anyone who has the connector | Only the tools their role grants, at server level and at tool level |
| Someone leaves | Revoke in every system and every client | One switch in your identity provider |
| Who keeps it running | Your team, from now on | Us |
Monday morning
A controller joins on Monday. She opens Claude, signs in with her company account, and the fourteen tools her role grants are simply there. She installed nothing, she filed no ticket, and she can reach no record she isn't allowed to see.
Who does the work
Nobody at your company writes code for this. You point us at a system, name the tools in your own words and mark the personal fields. That is configuration, not development.
Build it yourself if you have a platform team that wants to own an internal MCP gateway. Done properly, that means per-user OAuth, role resolution against your identity provider, policy per tool, masking on the way back and one audit trail across every assistant. At that point you are not building a connector, you are building a product. You can also have it on Friday.
How this compares.
| Loopthink | Microsoft SQL MCP Server | AnythingMCP | Composio | Anthropic MCP Tunnels | |
|---|---|---|---|---|---|
| Reaches systems not exposed to the internet | Yes | No | Not documented | Enterprise only | Yes, but not in claude.ai |
| Field-level masking before the model sees data | Yes | No | No | No | No |
| Runs as the person who asked | Yes | Partly | Not documented | Partly | No |
| One audit trail across Claude and ChatGPT | Yes | Per server | Per gateway | — | Tunnel events only |
| Your own connector domain and logo | Yes | — | No | No | — |
| Somebody builds it for you | Yes, €990 | No | No | No | No |
| Entry price | Free, then from €49 | Free | €19 | Free | Free, research preview |
Connecting a single database, and you have a developer team? Microsoft's SQL MCP Server is free, production-ready and runs on-premises. It is the right choice, and we don't compete with it. We govern it, alongside everything else you connect.
Watch ChatGPT reach an internal system.
No slides: we build an MCP server on an internal database, mask the sensitive fields, connect it to ChatGPT, and follow the request through the gateway.
Press play to load the video. Until then, nothing is requested from YouTube or any other Google server.
No new app. Just the AI they already open.
Your connector shows up inside the assistants your teams already use, and there's a sovereign UI for everyone else. Built on MCP, so every system that offers an MCP source is compatible out of the box.
ChatGPT
Your connector appears as a native ChatGPT connector, with governed access to the systems behind it.
Claude
The same connector in Claude's enterprise connector list, with per-user permissions and audit.
Langdock
EUA fully European stack: from interface to data path, sovereign end to end.
Open WebUI
Self-hostGoverned access in your self-hosted UI, ideal for air-gapped, regulated setups.
One policy across every client. Same access rules, DLP and audit everywhere. Add a client and it inherits the rules that are already in place.
Ask your systems. In plain language.
Your systems hold the answers, but only a few power users know how to get them. Loopthink puts a governed chat layer on top, so anyone can just ask.
Every question runs through the same gateway: the user's own permissions, full audit, DLP and EU-sovereign hosting.
Run the gateway our way, or entirely yours.
Same platform, same features. Start managed in the EU, or deploy fully on-prem for air-gapped and highest-regulation environments.
Loopthink Cloud
Fastest start- EU-hosted, fully managed. We run and update the gateway, audit-ready from day one.
- Fastest path to live. Ideal for most teams: no infrastructure to stand up, live in weeks.
- Runner stays yours. It runs inside your network. System credentials never reach the cloud.
Loopthink On-Prem
Maximum sovereignty- Your environment, entirely. The gateway runs in your cloud or your data center. Nothing routes through ours.
- Built for the strictest workloads. Air-gapped, BSI C5 and highest-regulation environments.
- Same feature set as Cloud. No capability trade-off for choosing sovereignty.
- No direct access to customer systems. Credentials remain in the Runner inside your network. The gateway never needs to see them.
- PII stays under your control. Masking happens in the Runner, before anything leaves your network.
- EU-hosted gateway, or hosted entirely by you. Run it in the EU under our operation, or in your own environment for zero third-party cloud dependency.
Free is where you start. Move to your own Runner the day your security team asks where the credentials live.
Every provider can now reach your network. None of them masks the column, none of them knows which of your people asked, and none of them keeps the log.
Shadow AI is already inside your systems.
MCP made connecting any AI to your data trivial. Convenient for employees. A board-level risk for you.
Every tool you use now ships its own MCP server. n8n, Make, Microsoft, your SaaS vendors. None of them ships the identity in front of it: one shared token for everyone, no record of who asked, no masking before the data leaves.
Shadow AI everywhere
Employees wire agents into internal systems without SSO, logging or approval, creating far more connections than IT expects.
Data leaves your control
Plaintext credentials, over-permissioned agents, and PII flowing to US AI providers under the CLOUD Act.
Compliance blocks rollout
EU AI Act, DORA and BSI C5 apply the moment an agent goes live. Without audit trails, legal says no.
of agentic AI projects will be cancelled by 2027, on weak governance and unclear ROI. Gartner.
Governing MCP Access to Enterprise Systems
Gateway architecture, field-level masking, and deployment topologies, written for architects, security engineers and CISOs. Deliberately explicit about what a gateway does not solve.
- Compliance mapped to controls. GDPR, EU AI Act, DORA and BSI C5 requirements against concrete technical and organizational controls.
- Vendor-neutral. The same policies across models and cloud providers, without locking governance into a single hyperscaler.
MCP authentication, in depth.
- August 26, 2026Do you need an MCP OAuth proxy? Five options comparedA remote connector wants an authorization server, most MCP servers have none. Five ways to put one in front, compared honestly, including the free ones.→
- August 25, 2026MCP authentication for claude.ai: what actually worksThe difference is not Desktop versus web, and that misreading costs hours. What a remote connector really requires, the four ways it fails, and five approaches compared.→
- August 24, 2026Your n8n workflow as a Claude connector: the complete setupSince June 2026 the MCP trigger speaks OAuth itself, and n8n’s own docs still say it does not. Three paths, when each one fits, and the tunnel settings that decide it.→
The questions that come up before the pilot.
- Who in our company has to add the connector?
- On Claude Team and Enterprise plans, only an Owner can add a custom connector. We hand you a prefilled link you can forward, plus a one-page note for your security team. On Free, Pro and Max you can add it yourself.
- Can't we just build our own MCP server?
- For one reachable database, yes, and the open source servers are good. What doesn't come with them is per-user identity into the target system, permissions at tool level, field masking, one audit trail across assistants, and somebody who patches it. That is the part we sell, and an existing MCP server keeps working: point it at us and it inherits all of it.
- Do we need developers?
- No. You connect a documented API yourself, or we build the connector for €990 in five working days. Naming the tools, setting the role matrix and marking personal fields is configuration, not development. What you do need is somebody who can decide which role sees what.
- Where do our system credentials live?
- In the Runner, inside your network. The gateway never needs to see them.
- How many tools will our people get?
- A curated set, typically under fifteen tools per connected system, named in your own business language, read-only by default. What one person sees is smaller again, because they only ever get the tools their role grants, not the catalogue. Read and write are always separate tools, and we deliberately do not expose every table.
- What happens after the five days?
- The connector is yours and running. You are on your monthly plan, and you can add systems yourself or book our team by the day.
- Do we need a Claude or ChatGPT contract?
- Yes, on their side. We govern the access, we don't resell the assistant.
- What does Loopthink not solve?
- Bad data stays bad data, and a governed connector is not a data warehouse. The whitepaper is deliberately explicit about the limits.
Bring your AI strategy under control.
Tell us where you are, and we'll set up a focused briefing for your industry and compliance needs.
- 30-min briefing: assess your setup, map compliance gaps, see if we fit. No pitch
- For IT and application leads, CIOs, CISOs and CDOs
- Your details stay in the EU
Your teams already use AI.
Give them one connector.
Yours, branded and governed, with your internal systems behind it and nothing exposed to the internet.
Get your AI strategy on track, EU-compliant and board-ready.