RELATED 01
Ruflo
Agent swarms with shared vector memory and task routing for Claude Code and other MCP clients
81kGitHub stars
87k/wknpm downloads
89popularity score
OmniRoute
Lets an assistant run an OmniRoute model gateway: routing combos, quotas, costs, cache, compression, memory
By Cheaper Inference TypeScript Local · stdio AI and ML
109 popularity score ★ 85k ↓ 97k/wk
Connect your client
Local (npm)
Add in one click
Opens the app, which asks before it adds the server.
Or add it by command or config
Claude CodeTerminal
claude mcp add omniroute -- npx -y omniroute --mcpClaude Desktopclaude_desktop_config.json
{
"mcpServers": {
"omniroute": {
"command": "npx",
"args": [
"-y",
"omniroute",
"--mcp"
]
}
}
}Cursor~/.cursor/mcp.json
{
"mcpServers": {
"omniroute": {
"command": "npx",
"args": [
"-y",
"omniroute",
"--mcp"
]
}
}
}VS Code.vscode/mcp.json
{
"servers": {
"omniroute": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"omniroute",
"--mcp"
]
}
}
}Codex~/.codex/config.toml
[mcp_servers.omniroute]
command = "npx"
args = ["-y", "omniroute", "--mcp"]Gemini CLITerminal
gemini mcp add --scope user omniroute npx -- -y omniroute --mcpWhat it does
OmniRoute is an MIT-licensed, self-hosted gateway that puts hundreds of model providers (paid, API-key and free tiers) behind one OpenAI-, Claude- and Gemini-compatible endpoint on your machine, with automatic fallback, routing strategies, token compression, caching, memory and guardrails. Its built-in MCP server lets an assistant operate that gateway.
The documentation lists 110 tools. Core tools report gateway health, quotas and costs, list and create routing combos, switch or test them, simulate and explain routing decisions, send a chat completion through the router, and search or fetch the web through configured providers. Others manage the response cache, compression engines and a content store, a memory store, skills, a free-proxy pool, and context sources such as Notion and Obsidian.
Install OmniRoute with npm install -g omniroute (or Docker) and run omniroute --mcp for stdio clients such as Cursor or Cline. A running OmniRoute also serves MCP over HTTP at http://localhost:20128/api/mcp/stream (and SSE at /api/mcp/sse) once enabled with omniroute mcp enable; Claude Code can add that URL directly. By default those endpoints answer only on loopback; remote clients need an API key with the manage scope.
What it needs
| Runtime | Node.js |
|---|
Tools 22
omniroute_get_health- Uptime, memory, circuit breakers, rate limits and cache stats.
omniroute_list_combos- List configured routing combos and their strategies.
omniroute_switch_combo- Activate or deactivate a combo.
omniroute_create_combo- Create a validated combo.
omniroute_check_quota- Quota used and remaining, reset time and token health.
omniroute_route_request- Send a chat completion through OmniRoute routing.
omniroute_cost_report- Cost report by session, day, week or month.
omniroute_list_models_catalog- Model catalog with capabilities, status and pricing.
omniroute_web_search- Web search through the configured search providers.
omniroute_web_fetch- Fetch web content through the configured fetch providers.
omniroute_simulate_route- Dry-run routing simulation with fallback tree.
omniroute_explain_route- Explain why a request was routed to a provider.
omniroute_set_budget_guard- Set a session budget with degrade, block or alert action.
omniroute_test_combo- Live test of every provider in a combo with a real upstream call.
omniroute_cache_flush- Flush the cache globally or by signature or model.
omniroute_compression_configure- Configure compression mode, threshold and target ratio.
omniroute_memory_search- Search stored memories.
omniroute_memory_add- Add a memory entry.
omniroute_skills_execute- Execute a registered skill.
omniroute_oneproxy_fetch- Fetch free proxies from the 1proxy marketplace.
notion_search- Full-text search across Notion pages and databases.
notion_append_blocks- Append blocks to a Notion page or block.