Use Provely where you already work
Connect Provely to the chat app you use, and check a prompt without leaving the conversation.
Your connection address
https://provely.app/api/mcp/mcpClaude
How to connect
- Open Settings, then Connectors
- Choose Add custom connector
- Paste the address above and save
- Choose Connect, then sign in to Provely
On the free plan you can keep one custom connector at a time.
Claude Code
claude mcp add --transport http provely https://provely.app/api/mcp/mcpHow to connect
- Run the command above in your terminal
- Type /mcp in a session
- Sign in to Provely when the browser opens
If you have already connected Claude, Claude Code picks it up on its own. You only need one of the two.
ChatGPT
How to connect
- Open Settings, then Plugins, and turn on Developer mode
- Go back to Settings, then Plugins, and press the plus button
- Paste the address above and leave everything else as it comes
- Choose sign-in with your account, then sign in to Provely
Needs a paid plan on the web: Plus, Pro, Business, Enterprise or Edu. If the Developer mode switch is not under Plugins, look under Settings, then Security and login.
Codex
How to connect
- Open ~/.codex/config.toml, or .codex/config.toml inside a project
- Add Provely, as shown below
- Run codex mcp login provely in your terminal
- Sign in to Provely when the browser opens
Add this to the file
[mcp_servers.provely]
url = "https://provely.app/api/mcp/mcp"
auth = "oauth"Codex reads the file when a session starts, so begin a new one after you save. The sign-in line only works once the entry is in the file.
Gemini CLI
gemini mcp add --transport http provely https://provely.app/api/mcp/mcpHow to connect
- Run the command above in your terminal
- Sign in to Provely when Gemini asks
- If it does not ask, type /mcp auth provely
Gemini handles the sign-in itself. There is nothing to paste.
VS Code
How to connect
- Run MCP: Add Server from the Command Palette
- Choose HTTP and paste the address above
- Name it provely, then choose Global to have it everywhere
- Let VS Code open your browser, and sign in to Provely
Add this to the file
"provely": {
"type": "http",
"url": "https://provely.app/api/mcp/mcp"
}Needs VS Code 1.102 or newer, where this became built in. There is no extension to install. To do it by hand instead, run MCP: Open User Configuration and add the entry above.
Cursor
How to connect
- Open ~/.cursor/mcp.json, or .cursor/mcp.json inside a project
- Add Provely to the servers list, as shown below
Add this to the file
"provely": {
"url": "https://provely.app/api/mcp/mcp"
}Cursor documents the config file, so that is what we describe. If it asks you to sign in, that is expected.
Windsurf
How to connect
- Open ~/.codeium/windsurf/mcp_config.json
- Add Provely to the servers list, as shown below
- Sign in to Provely when it asks
Add this to the file
"provely": {
"serverUrl": "https://provely.app/api/mcp/mcp"
}Windsurf is being renamed to Devin Desktop, so a newer build may label its settings differently. The file path is unchanged.
- Check a prompt
check_prompt
Check a prompt before you rely on it, and get it back improved in the same call. - Fix a chat that went wrong
fix_chat
Reads a conversation that went wrong and says where it turned. - Read the Context
get_context
Read the facts the user keeps in their Provely Context: the company they work for, who they write for, the constraints they apply every time. - Set up a new Context
set_up_context
Start a new Provely Context by answering questions on the website, with no need to send facts one by one. - Read the team prompts
get_team_prompts
Read the prompts the user's Provely team has proved and saved, and pull one back in full. - Get a test result
get_test_result
Fetch work that was still running when a Provely tool had to answer. - Give feedback on a result
give_feedback
Save the user's feedback on a Provely result: a prompt check, a rescued conversation, an answer check or a prompt batch. - Prove a prompt
prove_prompt
Apply the fixes the user chose to a checked prompt, then prove the result. - Save a prompt to the team
save_prompt_to_team
Save a prompt the user has proved into their Provely team, so their colleagues can start from what already works. - Save facts to the Context
sync_context
Save facts you already know about the user into their Provely Context, so later checks are read against what is actually true for them. - Test on the other models
test_other_models
Run an already-checked prompt on the frontier models it has not met yet, to see whether it holds up beyond the one model the check tested. - Verify an AI response
verify_response
Read an AI response for unsupported, invented or overstated claims, against the context the user provides. - Start a batch of prompts
start_prompt_batch
Turn a dump of work into a batch of prompts, one per job, the way the Provely website's multi-prompting does. - Answer a batch and prepare its prompts
answer_prompt_batch
Answer a prompt batch's questions and prepare its prompts. - Finish a batch of prompts
finish_prompt_batch
Apply the fixes the user kept and hand back the finished prompts, each with the model to run it on and why, plus how much time and how many retries the batch is estimated to save.