How to Connect PhantomBuster to Meta Muse via MCP

💡 No install? Use cloud.anythingmcp.com directly. Sign in, click Connectors → PhantomBuster, paste your credentials, mint an MCP API key and you're done. No Docker, no git clone, no local server.

PhantomBuster + Meta Muse

Drive a PhantomBuster workspace from any AI agent: list and read your agents (Phantoms), launch, schedule and stop them, follow their runs and read the results, edit their input and schedule, work with leads lists and stored leads, and check plan usage. API v2 with an API key; deletes and the emergency unschedule install switched off.

Prerequisites

See the full setup instructions baked into the connector (visible in the in-app store when you select the connector). The required environment variables for this connector are:

PHANTOMBUSTER_API_KEY

Step 1 — Get credentials

Setting up takes about 2 minutes. You need a PhantomBuster workspace with the Phantoms you want the AI to run.

1. In PhantomBuster open Workspace settings (https://phantombuster.com/workspace-settings, also in the navbar menu), go to API Keys and add a key. It is shown only once: copy it.
2. Paste it into PHANTOMBUSTER_API_KEY. It is sent in the X-Phantombuster-Key header and gives full access to that one workspace.
3. Install the connector and run phantombuster_get_org: it shows the workspace, its plan and time zone and confirms the key.

What it reads: the workspace with plan, time zone and daily limits; plan usage (execution time, email, AI, captcha and SERP credits and storage, today and this month, with the next reset); the runs executing right now; CSV exports of usage per agent and per run over the last N days (at most about 6 months); the dashboard folders. Your agents with script, schedule, last result and saved input, the deleted agents, and the scripts with their manifest (the input fields an agent of that script accepts). The runs (containers) of an agent with status, times and exit code, their console log and their result data. The leads lists of the workspace storage with their saved filter, and the leads matching a list (name, LinkedIn profile and company, job title, location, emails and the other stored fields).

What it writes: launches an agent now (phantombuster_launch_agent) or within the next N minutes (phantombuster_launch_agent_soon), stops a running agent (phantombuster_abort_agent, with soft_abort the Phantom finishes its current step and saves what it has), changes an agent's name, saved input, schedule (launch_type, repeated_launch_preset), time limit and notifications or creates a new agent from a script (phantombuster_save_agent, which never starts a run), creates or changes leads lists and stores or updates single leads.

Launching is real: a run acts on the social accounts connected in PhantomBuster (it visits profiles, sends connection requests or messages) and uses the plan's execution time. Before a launch the AI shows you the agent, its input and the volume and waits for your yes. Too many actions a day can get a LinkedIn account restricted, so keep the daily limits the Phantom suggests. Session cookies and connected accounts are set up in PhantomBuster itself; never paste a cookie into the chat.

Input for one run: argument replaces the saved input for that run (and is stored too with save_argument); bonus_argument merges a few fields over the saved input for one run only. The field names come from the script manifest (phantombuster_get_agent with with_manifest: "true"), for example spreadsheetUrl or numberOfProfiles.

Following a run: the launch returns a containerId. The AI then polls phantombuster_get_agent_output about once a minute, reading only the new log lines with from_output_pos and prev_container_id, until the status is finished. The data of a run comes from phantombuster_get_container_result; the files are also at https://phantombuster.s3.amazonaws.com/<orgS3Folder>/<s3Folder>/result.csv (and result.json), with both folders from phantombuster_get_agent. Two agents can share a name, so the AI always works with agent ids.

Leads lists: a list is a saved filter over the stored leads and updates itself as Phantoms store new leads. A filter is {"and": [ ... ]} of clauses such as {"filter": {"linkedin_job_title": {"entity": "lead", "operator": "contains", "valueToCompare": "CTO"}}}, nested and / or, or {"__global_search__": "text"}. Lead fields are snake_case (linkedin_job_title, location, professional_emails), company fields camelCase with "entity": "company" (industriesV2, employeeCount). The leads storage tools use PhantomBuster's beta API.

Switched off at install: phantombuster_delete_leads (cannot be undone), phantombuster_delete_lead_list (the leads stay in storage), phantombuster_delete_agent (its past runs and result files are no longer shown) and phantombuster_unschedule_all_agents (the emergency stop that turns every agent to manual launch). A workspace admin turns them on when the team needs them.

Data formats: ids are strings of digits. Dates are Unix timestamps in milliseconds. Run status is starting, running, finished, launch error or never launched; exitCode 0 with endType finished means success. Usage values are seconds of execution time and counts of credits.

Paging: phantombuster_list_containers returns limit runs, newest first; for older ones the AI passes the endedAt of the last run as before_ended_at. phantombuster_list_leads_in_list pages with page_size and offset (include_total_count gives the total). The other lists return everything at once. The list of PhantomBuster's store scripts is several MB, so the AI filters it with script_ids.

Common errors: 401 "API key not found" means a wrong or revoked key. 400 means a missing or wrong parameter, or a launch argument that is too big. 404 is an unknown id. A launch fails when the plan's execution time is used up or too many runs are already going in parallel (maxParallelism).

The connector does not connect social accounts or handle session cookies, does not edit Phantom code (scripts are read-only) and does not change the plan.

Step 2 — Install the adapter

curl -fsSL https://raw.githubusercontent.com/HelpCode-ai/anythingmcp/main/docker-compose.quickstart.yml -o docker-compose.yml
printf 'JWT_SECRET=%s\nENCRYPTION_KEY=%s\n' "$(openssl rand -hex 32)" "$(openssl rand -hex 32)" > .env
docker compose up -d

Step 3 — Add the connector in Muse

Your server URL: in AnythingMCP, open MCP Servers → the server this connector is on, and copy its URL (https://cloud.anythingmcp.com/mcp/…). Use it wherever this guide shows YOUR_SERVER_ID.

Muse adds remote MCP servers as custom connectors and signs in to them with OAuth.

1. In Muse, open Settings → Connectors and choose Add custom connector.
2. Paste the MCP server URL https://cloud.anythingmcp.com/mcp/YOUR_SERVER_ID.
3. Muse opens the AnythingMCP sign-in page: sign in and approve access.
4. In a chat, ask Muse for something that needs the PhantomBuster tools.

You can also set it up from a chat, on the web, in the iOS and Android apps or in WhatsApp. Send Muse this message:

Create a Custom Connector for a new remote MCP server, then connect to it:
Name: AnythingMCP
Transport: remote streamable HTTP
URL: https://cloud.anythingmcp.com/mcp/YOUR_SERVER_ID
Auth: OAuth

Self-hosting? Use your instance's public HTTPS address instead of https://cloud.anythingmcp.com. Muse connects from Meta's cloud, so localhost does not work.

Available tools

Example prompts

• "Which of my Phantoms ran in the last 24 hours, and did any of them end with an error?"
• "Launch my Sales Navigator Search Export agent with the search URL I give you, limited to 100 profiles. Show me the input first."
• "How much execution time and how many email credits are left on our plan this month?"
• "Create a leads list of all stored leads whose job title contains CTO and who are located in Germany."
• "Change the schedule of my LinkedIn Profile Scraper to run once a day, and tell me when it runs next."

FAQ

Where is Muse available? Meta currently offers Muse in the US and Canada, on the web, in the iOS and Android apps and in WhatsApp.

Does Meta review this connector? No. Meta doesn't review custom connectors. Muse sees only the tools you switched on for this MCP server in AnythingMCP, and you can disconnect it anytime under Settings → Connectors.

Next steps

• Connector landing page
• Connect to Claude
• Connect to ChatGPT
• Connect to Cursor
• Connect to Gemini
• Connect to Microsoft Copilot