The ultimate LLM gateway for agents. Route models, enforce spend limits, and debug traces. Get the official MCP server via npm or our private registry.
Connect your AI to the SpiderGate MCP server.
SPIDERIQ_MCP_MODE=facade SPIDERIQ_FORMAT=yaml npx -y @spideriq/mcp@latest
Give your agent the knowledge to manage the gateway and pick the best models.
# Install CLI first: npm i -g @spideriq/cli
spideriq skills add @spideriq/gateway-skills
Matched set as of 2026-08-28 — core 1.74.0 · mcp-gate 1.10.0 · mcp 1.89.0 · cli 1.71.0 · gateway-skills 0.11.0.
gate_capacity and gate_flow need mcp-gate 1.10.0 or later. On an older pin they do not exist and nothing tells you so.
Secure your connection with your SpiderGate brand token.
SPIDERIQ_TOKEN=spideriq_pat_your_token
Ask your agent to route tasks, check costs, or read traces.
"Route this task to the best coding model under $1/M tokens"
Every tool follows the SpiderGate isolation architecture, ensuring secure multi-tenant capabilities.
Route prompts, embeddings, and strict JSON output across top providers. Use aliases for fallback handling, or pin direct models for strict control.
12 toolsLet your agent pick the right model for a job. Filter by category, enforce budget caps, and read machine and human reviews before picking.
10 toolsWhat your traffic actually delivered, not just what returned 200. A four-state outcome split over any calendar range, key pressure with a verdict that can refuse, the alias-to-model cross-tab, and one request end to end.
7 toolsA massive creative suite. Save prompts to a shared gallery. Group work into projects with reference files. Generate images, video, and voice that yield durable URLs.
6 toolsA safe vault for Bring-Your-Own-Key. Securely save provider credentials. Let third parties add keys safely via the Invite Contributor flow. Set distinct agent tokens with hard token budgets and rate limits.
30+ toolsOne conversation. Six tool calls. A fully deployed page — live on your custom domain in under 30 seconds.
Run the gateway using the CLI, connect fast via the MCP server, or drop native skills right into your agent's code.
Agents can use top models safely. They never see the raw API keys. You can also harvest keys from clients using a secure email link. This keeps your vault locked down.
The gateway sends data back in YAML and Markdown instead of heavy JSON. This cuts your token costs, frees up context limits, and makes agent loops run faster.
Stop hardcoding links. Agents can search the catalog to find and pick the best models. We score models with real machine and human tests so your agent always wins.
Keep all your prompts in one place. Agents can load and run shared prompts on the fly. Turn single scripts into one shared team brain.