Connect Guide

Dashboard → Connect is the single place to copy MCP configuration for your deployment. SmartGate does not auto-register with LLM apps — you paste URL + API key into each host.

Before you start

• An active SmartGate account and API key (Quick Start)
• An MCP client that supports remote Streamable HTTP with custom Bearer headers

Connect workflow

1. Open Dashboard → Connect.
2. Select your platform tab (Cursor, Claude Desktop, Windsurf, Hermes, OpenClaw, Generic).
3. Copy the JSON or CLI command shown.
4. Merge into your host config file (Connect shows the path).
5. Restart the host app (or use Refresh MCP where supported).

Connect always reflects your current deployment URL: POST {site}/api/mcp (Streamable HTTP, stateless).

Playground is not MCP

Dashboard → Playground runs tests through the REST API. It does not use your MCP Connect config.

Per-host reference

Official host docs:

• Cursor MCP
• Claude Code MCP
• Windsurf Cascade MCP
• OpenClaw MCP

SmartGate uses MCP 2025-03-26 Streamable HTTP only. Legacy HTTP+SSE dual-endpoint transport has been removed.

OpenClaw tips

• Use transport: "streamable-http" explicitly.
• SmartGate normalizes common JSON-RPC quirks server-side.

Supported clients

Any client with remote Streamable HTTP + Bearer headers, including Cursor, Claude Desktop, Windsurf, Hermes, OpenClaw, Cline, Zed, and custom agents.

Troubleshooting

Full protocol details: MCP Endpoint.

Why Connect is the source of truth

Hosts differ in file paths and JSON shapes, but they agree on the transport SmartGate supports: Streamable HTTP to POST {site}/api/mcp with a Bearer API key. Dashboard Connect always renders the URL for your deployment, so you should copy from Connect rather than hand-typing an example from a blog post. After you paste config, restart or refresh MCP so the host re-runs initialize and tools/list.

SmartGate does not auto-register into Cursor, Claude Desktop, Windsurf, or OpenClaw. That is intentional: API keys stay under your control, and each host keeps its own MCP server map. If tools appear empty, complete the handshake first; if you see 401, fix the Bearer header; if you see 405 on GET, switch to POST; if you hit legacy /messages, update to /api/mcp.

Playground vs MCP path

Playground exercises REST flows for evaluation. It does not load your MCP Connect JSON. When you validate an agent integration, use the host with Connect config. When you validate a dashboard preset, use Playground while signed in. Mixing the two is the most common cause of "it works in Playground but not in Cursor" reports.

Host checklist

1. Create an API key with the minimum role your team needs (Teams).
2. Open Connect and select the matching tab.
3. Paste into the path Connect shows; do not rename tool prefixes unless your host requires it.
4. Confirm Streamable HTTP (not legacy SSE dual-endpoint).
5. Call a cheap tool (for example a small smart_search or smart_budget_guard check) and confirm an Activity Log row (Logs).
6. Only then wire long smart_pipe research templates.

Positioning reminder

SmartGate is an MCP-native intelligence layer—five capabilities, seven primitives—not a host LLM. Connect wires the Agent Interface; human Web scenes stay on marketing and dashboard paths without exposing MCP JSON to end users who never need it.

Practical adoption notes

Treat SmartGate as shared infrastructure: one gateway URL, team-scoped keys, and audit on by default. Prefer pipe templates for repeatable research paths so Activity Logs show a coherent chain instead of ad-hoc tool spam. Keep host-model prompts focused on decisions; let Research and Context primitives prepare evidence. When FinOps asks for savings, export Usage Reports alongside the gateway benchmark explanation—never present Est. USD saved as a vendor LLM invoice. When security asks for evidence, export an audit window and show retention tied to the current plan catalog.

If you are migrating from scripts, replace raw HTTP fetch helpers first, then add dedup and compression, then budget guard. If you are migrating from a model gateway mindset, stop expecting chat completions on /api/mcp and wire MCP hosts instead. If you are writing human-facing product copy, lead with Smart Research / Smart Decision outcomes; keep MCP JSON on builder paths. These habits keep the site and the product aligned with the Positioning Bible: intelligence layer between AI and the world, not an LLM proxy.

Related reading

• Overview for the capability map and plan table
• Connect Guide for host JSON and troubleshooting
• Tools Reference for when to search, fetch, compress, or pipe
• Activity Logs for audit, trace, and export
• Pricing for fair-use limits, Gainshare, and plan ceilings
• Token Control, Traffic Customs, and Audit and Compliance for pillar narratives aligned with this docs set

Keep English-first public docs consistent with the Positioning Bible: SmartGate is an MCP-native algorithmic gateway and intelligence layer—not a chatbot, ChatGPT alternative, AI search engine, LLM proxy, or model router. Humans research and decide through Web experiences; agents connect through MCP/API; both share the same runtime primitives and team-scoped governance.

One more operational reminder

Document your team naming scheme, key rotation cadence, and who owns Gainshare questions before you scale agent count. Prefer bounded Playground or single-tool MCP checks when validating Connect JSON. Prefer audited smart_pipe templates for recurring research once the handshake works. Prefer exporting Activity Logs before plan downgrades that shorten retention. These habits cost little and prevent the two failure modes we see most: empty tool lists from incomplete MCP initialize, and missing audit evidence when a reviewer asks what an agent fetched last week.