Applied AI
MCP servers and Claude connectors: putting your product inside AI assistants
An MCP server exposes a product’s actions and data to AI assistants through the Model Context Protocol, the open standard Claude, ChatGPT, Cursor, VS Code and other AI tools use to call external tools. A production MCP server needs the same things as a public API, plus OAuth sign-in that works from inside the assistant, tool descriptions a model can use correctly, and billing that counts usage across every way in. Hexploits builds and runs them, and took its own PII Purge connector through review into the Claude Directory.
A named engineer replies within one working day. A written scope and an indicative price within two.
Who this is for
- SaaS companies whose customers ask to use the product from Claude, ChatGPT or Copilot.
- Product teams that want a listing in the Claude Directory or another AI connector marketplace.
- Businesses that want staff assistants to reach internal systems under existing permissions, rather than by copy and paste.
- API-first products that need usage from AI assistants metered and billed alongside the API.
What you get
Deliverables, not slogans. Each one appears in the statement of work.
- A remote MCP server over Streamable HTTP, built on your existing API or alongside it.
- OAuth sign-in with discovery from the authentication challenge, so a user connects from the assistant in a browser window, plus API keys for developer clients.
- Tool design: names, descriptions, inputs and outputs written and tested so models choose and call them correctly.
- Safe handling of files, URLs and large inputs, including size limits and refusal of private-network addresses.
- Usage metering and tiered billing through Stripe, shared with the website and API, with failed calls left uncharged.
- Submission material and review fixes for the Claude Directory, and setup guides for Claude, Claude Code and other MCP clients.
Our engineers work across the major languages, frameworks and cloud platforms. We build on the stack you already run, with technology choices explained in writing before work begins.
How it is delivered
The same four stages as every Hexploits engagement, applied to this capability.
Stage 1
Choose the tools
We agree which actions an assistant should be able to take, which need confirmation, and which stay out of reach. Output: a tool list with inputs, outputs and limits.
Stage 2
Build
The MCP server, authentication and metering, tested against Claude, Claude Code and other clients with real prompts, not only protocol tests.
Stage 3
List
Directory submission, documentation and setup guides, and changes from review until the connector is approved.
Stage 4
Operate
Monitoring of calls, errors and usage, and updates as the protocol and the assistants that implement it change.
Built and listed by us
PII Purge in the Claude Directory
PII Purge, our PII redaction product, is a remote MCP server listed in the Claude Directory. Anyone can add it to Claude or Claude Code, sign in with OAuth, and redact text and PDFs from a conversation. Usage is metered through Stripe in tiers shared with the website and REST API, and every LLM step runs on Hexploits Cloud in the EU.
How success is measured
Every engagement agrees its measures and the measurement period in writing before work starts.
- Successful tool calls as a share of all calls, by client.
- Time from a user adding the connector to their first successful call.
- Usage and revenue from AI assistants, reported separately from the website and API.
Proof
Case studies with numbers, and reviews linked to Google where they were left there.
PII Purge · Data protection
99.7% of personal data redacted and nothing retained: GDPR document redaction for CVs, subject access requests and AI pipelines
swarmd.ai · Software vendor
Policy on every agent call and nine services in production: an enterprise AI control plane for a UK software vendor
Questions we get asked
What is the Model Context Protocol?
How do we get our product into the Claude Directory?
Do we need to rebuild our API?
How is usage from AI assistants billed?
Is it safe to let an assistant call our systems?
Related
Sectors where this is most often needed
Reading
ACP: The Commerce Layer for AI Agents
Customers now buy inside AI conversations. The Agentic Commerce Protocol from Stripe and OpenAI lets an AI agent run your checkout while you stay merchant of record. What it costs to adopt, who is already selling through it, and what a retailer should do now.
Decision APIs: Fast, Cheap AI Judgements and Where They Fall Short
Decision APIs pick from answers you define instead of writing text. What OpenAI’s Decisions API, Jev and rivals can and can’t do, with prices and benchmarks.
More in applied ai
Internal AI assistants and copilots for UK businesses
EU AI Act readiness for UK businesses
AI document and email processing
AI decision support and forecasting
Agentic AI workflows with governance
AI sovereignty: private and EU-hosted AI
AI governance and ISO 42001
AI in payments, credit and fraud under the EU AI Act
AI for recruitment: sourcing, screening and matching automation
EU AI Act for recruitment and HR: high-risk hiring AI explained
AI governance readiness for recruitment and HR businesses
Next step
Request a proposal.
Tell us about the system and the sector. A named engineer replies within one working day. A written scope and an indicative price within two working days of a short scoping call.