Self-hosted MCP gateway with PII rehydration and two-layer policy hooks.
Pseudonymize emails, names, IDs before the LLM sees them — rehydrate for the actual tool call. YAML hooks enforce policy at company and per-user scope, hot-reloaded. One endpoint for Claude, Codex, ChatGPT, Gemini.
Why mcpgate?
Chains across your tool stack in one prompt
The AI knows which Jira ticket links to which MR links to which Slack thread. One prompt, no tab switching, no copy-paste.
# "What's blocking the Q2 release?"
jira: PBE-2674 → blocked by PBE-2701
gitlab: MR !891 awaiting review
slack: thread in #release, 14:32
# Answer assembled. No copy-paste. Markdown to ADF, project templates, link-back — automatic
Pre-hooks apply Jira project templates, convert Markdown to Atlassian Document Format, link the ticket back to the Slack thread. The AI writes prose; the gateway handles the schema.
Compliance: audit, sanitization, exfiltration alerts
PII pseudonymized before the LLM sees it, rehydrated for the actual tool call. The audit log captures every action plus byte volume — so a sudden 50 MB outbound spike at 3 am is visible, not invisible.
- Pass-through architecture — stores nothing in transit
- Per-user request volume with Slack alerts on threshold breach
- Destructive actions require
confirmed=true - Encrypted pseudonym mapping with 24h TTL
Works with every MCP‑compatible AI
Built for Claude Code first. One admin configures it at claude.ai — the entire team gets access in every session. No individual setup.
Connect as an App. Available in every conversation with read/write separation and user consent per action.
Codex, Gemini CLI, or your own agents. MCP is an open standard — if it speaks MCP, it works.
Up and running in minutes
Deploy
$ docker run -d \
-v ./config:/config \
-p 8080:8080 \
mcpgate/mcpgate:latest Configure
services:
- slack
- jira
- google_workspace
- gitlab Connect your AI
mcpServers:
mcpgate:
url: "https://your-gateway/mcp"
token: "mcp_..." Read first
All posts →What flows through, what gets blocked, what gets logged
Audit, PII sanitization, throughput as a DLP lens, policy hooks — four layers stacked on the same audit trail. How they fit together, and the data-exfiltration signal you do not want to be missing.
Read →Dynamic action discovery isn’t novel. Doing it safely is.
Anthropic shipped Tool Search in November 2025. mcpproxy-go has BM25. Stacklok benchmarks 94% retrieval. What the gateway market converged on, and what mcpgate chose differently.
Read →mcpgate vs Obot: which self-hosted MCP gateway?
Side-by-side: where Obot is the better choice (maturity, MIT license, Kubernetes day one), where mcpgate is (PII pseudonymization, two-layer policy hooks).
Read →Try mcpgate now
Explore the full gateway with demo data. No signup, no install, no credit card.