Every AI agent, sharing the same MCP servers.
Set up a server once in Toolport and every agent below can use it. One toggle per agent, and Toolport writes its gateway entry into that agent's own config, next to anything already there, with your full server set showing up and lazy discovery keeping the token cost flat.
Chat & local model apps
-
Claude Desktop
Anthropic's desktop app, connected to every server through one gateway entry.
set up Claude Desktop -
LM Studio
local modelsLocal models with tool use, without pasting server configs into mcp.json by hand.
set up LM Studio -
Jan
local modelsThe open source local assistant, with your full server set one toggle away.
set up Jan -
BoltAI
The macOS AI chat app, sharing the same servers as your editor and CLI.
set up BoltAI -
Open WebUI
local modelsConnects natively over the gateway's HTTP/OpenAPI endpoint. No bridge, no mcpo.
set up Open WebUI -
AnythingLLM
local modelsThe all-in-one desktop assistant, sharing one governed server set.
set up AnythingLLM -
Witsy
local modelsThe open source desktop AI assistant, pointed at one gateway.
set up Witsy
Code editors & IDEs
-
Cursor
One gateway entry in ~/.cursor/mcp.json instead of a server list to maintain.
set up Cursor -
VS Code
Copilot agent mode with every server, from one entry in your user mcp.json.
set up VS Code -
Windsurf
Cascade reaches every server through one mcp_config.json entry.
set up Windsurf -
Zed
A context_servers entry written into settings.json without touching the rest of your config.
set up Zed -
Antigravity
Google's agent-first editor, with your servers shared from the same gateway.
set up Antigravity -
Kiro
AWS's agentic IDE, connected through one entry in its settings mcp.json.
set up Kiro
CLI agents & terminals
-
Claude Code
One global gateway entry in ~/.claude.json; every project inherits it.
set up Claude Code -
Codex
CLI and desktop app share ~/.codex/config.toml; Toolport writes the TOML for you.
set up Codex -
Gemini CLI
Google's terminal agent, wired to every server through settings.json.
set up Gemini CLI -
Goose
local modelsBlock's open source agent; Toolport writes the extensions entry in config.yaml.
set up Goose -
Amazon Q
Amazon Q Developer CLI, connected through ~/.aws/amazonq/mcp.json.
set up Amazon Q -
Warp
The agentic terminal, reading Toolport's gateway from its file-based MCP config.
set up Warp -
Hermes
Nous Research's agent, configured via the mcp_servers map in config.yaml.
set up Hermes -
Factory Droid
Factory's terminal agent, wired to every server through ~/.factory/mcp.json.
set up Factory Droid -
Crush
Charm's terminal agent, sharing the same server set as the rest of your tools.
set up Crush -
Amp
Sourcegraph's agent, connected through one entry in its settings file.
set up Amp -
OpenCode
The open source terminal agent, pointed at one gateway instead of many servers.
set up OpenCode -
Grok Build
xAI's terminal coding agent, wired in with one TOML table.
set up Grok Build -
GitHub Copilot CLI
Copilot in the terminal, sharing every server through one gateway entry.
set up GitHub Copilot CLI -
Qwen Code
Alibaba's terminal agent, connected through its settings.json.
set up Qwen Code -
Kimi Code
Moonshot's terminal agent, sharing your whole server set from one entry.
set up Kimi Code -
Pi
The minimal terminal coding harness, with every server one search away.
set up Pi -
Oh My Pi
The Pi fork with LSP and a debugger, connected the same one-click way.
set up Oh My Pi
Editor extensions
-
Cline
The VS Code agent extension, pointed at one gateway instead of N servers.
set up Cline -
Roo Code
Roo's whole agent team shares one server set through the gateway.
set up Roo Code -
Continue
The open source VS Code and JetBrains assistant, wired via config.yaml.
set up Continue -
Kilo Code
The open source VS Code agent, sharing one governed server set.
set up Kilo Code -
JetBrains Junie
JetBrains' agent across every IDE, wired in once at the user level.
set up JetBrains Junie
Using an agent that isn't listed? If it reads an mcpServers-style config, the
gateway entry usually works as a manual add, and we're quick to add first-class support.
Open an issue or ask on
Discord.