Connect Claude & Any MCP Client
One MCP endpoint gives your agent the whole FlyMy.AI platform: generative models, a full catalog of service connectors, cloud agents you can create and freeze into callable APIs, skills, files, and schedules. If a client speaks MCP over Streamable HTTP, it can use it.
https://mcp-agents.flymy.ai/mcp
30-second connect
In Claude Code, one command:
claude mcp add --transport http flymyai https://mcp-agents.flymy.ai/mcp --header "X-API-Key: YOUR_KEY"
Grab your key at app.flymy.ai/profile, replace YOUR_KEY, and ask Claude to whoami - if it answers with your account, you are connected.
GUI clients (Claude Desktop, claude.ai, ChatGPT) authenticate with an OAuth sign-in when you add the connector - no key to paste. CLI and config-file clients (Claude Code, Cursor, Codex, Windsurf) pass your API key in an X-API-Key header instead.
Per-client setup
- Claude Code
- Claude Desktop
- claude.ai
- ChatGPT
- Cursor
- Codex
- Windsurf
- Antigravity
claude mcp add --transport http flymyai https://mcp-agents.flymy.ai/mcp --header "X-API-Key: YOUR_KEY"
Verify with /mcp inside a Claude Code session - flymyai should show as connected with its tool list.
- Open Settings -> Connectors
- Click Add custom connector
- Paste
https://mcp-agents.flymy.ai/mcp - Sign in with your FlyMy.AI account when prompted
No API key needed - the OAuth sign-in links the connector to your account.
- Open Settings -> Connectors
- Click Add custom connector
- Paste
https://mcp-agents.flymy.ai/mcp - Sign in with your FlyMy.AI account when prompted
Once added, enable the connector in a chat and the tools become available to Claude.
- Open Settings -> Connectors
- Click Add custom connector
- Paste
https://mcp-agents.flymy.ai/mcp - Sign in with your FlyMy.AI account when prompted
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"flymyai": {
"url": "https://mcp-agents.flymy.ai/mcp",
"headers": {
"X-API-Key": "YOUR_KEY"
}
}
}
}
Add to ~/.codex/config.toml:
[mcp_servers.flymyai]
url = "https://mcp-agents.flymy.ai/mcp"
http_headers = { "X-API-Key" = "YOUR_KEY" }
Add the same mcpServers block to Windsurf's MCP config file:
{
"mcpServers": {
"flymyai": {
"url": "https://mcp-agents.flymy.ai/mcp",
"headers": {
"X-API-Key": "YOUR_KEY"
}
}
}
}
Open Antigravity's MCP settings and add the endpoint:
https://mcp-agents.flymy.ai/mcp
What your agent can do now
The gateway exposes about 54 tools. The list is dynamic - your client discovers it at runtime via tools/list, so treat what your client shows as the source of truth.
Run any model
recommend_model picks the best FlyMy.AI model for a task (text-to-image, video, and more), get_model_schema returns its exact parameters, run_model executes it. list_media_models, flymyai_models, and get_pricing round out discovery.
"Generate an image of a lighthouse at sunset in watercolor style."
Reach the whole connector catalog
search_tools finds the right connector action (Slack, Notion, GitHub, Google Workspace, CRMs, and more), execute_tool runs it. add_tool, update_tool, delete_tool, set_tool_config, provide_tool_config, and list_configured_tools manage your configured set. You can also plug in your own servers with add_mcp_server, connect_mcp_server, call_mcp_server, and the matching list / update / delete / disconnect tools.
"Post this release summary to Slack."
Create, run, and freeze cloud agents
create_agent, run_agent, get_run, and append_message cover the build-and-run loop, with list_agents, get_agent, update_agent, copy_agent, delete_agent, cancel_run, favorite_run, set_agent_public, set_effort, and suggest_schema alongside. Once a run does what you want, freeze_agent compiles it into a fixed instruction and run_frozen replays it - deterministic, cheaper, faster. schedule_agent puts a frozen compilation on a cron schedule (5-field expression plus an optional IANA timezone), and get_compilation, list_compilations, update_compilation manage versions. See Freeze & Re-run for the semantics.
"Freeze this as an API I can call from my backend."
Skills, files, billing, identity
- Skills:
create_skill,get_skill,list_skills,attach_skill,detach_skill- reusable instruction packs you attach to agents. - Files:
attach_files,list_agent_files,files_tree,download_file,delete_agent_file- give agents inputs and fetch their outputs. - Billing:
get_execution_price,list_execution_prices- the real billed cost of any run. - Identity:
whoami- which account the connection is using.
On sensitive actions the agent pauses the run and raises an approval request in the app at app.flymy.ai. The run continues after the owner approves.
Auth & billing
Everything runs under your account: GUI clients link it via the OAuth sign-in, CLI and config clients via the X-API-Key header. Keys live at app.flymy.ai/profile.
Billing is per-use - no subscription. Every run has a real billed cost you can query with get_execution_price (agent runs) or get_pricing (models). For scale, published real bills: a dictation note costs $0.031/run, a short video clip roughly $0.20-0.50 depending on model.
The key or signed-in account on the connection is the one billed. Do not commit config files containing YOUR_KEY to a shared repo - keep the key in a local, untracked file.
Next steps
- MCP Integrations hub - browse the full connector catalog
search_toolsdraws from - Generate media from Claude - the image and video cookbook for this connection
- Freeze & Re-run - how frozen compilations behave and when to cut a new version