ctm-multi-mcp
Read-only MCP server for extracting data across multiple CallTrackingMetrics sub-accounts using a single parent API key. Enables cross-account call data, reports, numbers, contacts, and configuration analysis for resellers and agencies via Claude.
README
ctm-multi-mcp — Multi-account CTM data extraction
A read-only Model Context Protocol (MCP) server for extracting data across multiple CallTrackingMetrics sub-accounts using a single parent/reseller API key.
Built for resellers and agencies managing multiple CTM accounts who need to pull call data, reports, number inventory, contacts, sources, and configuration into custom Claude bots.
How It Works
- Auto-discovery — On startup, the server uses the parent API key to list all sub-accounts automatically. No manual configuration needed.
- Filtering — Optionally limit which accounts are queried via env var or a JSON config file with optional labels.
- Cross-account tools — Every tool runs across all (or selected) accounts and returns results with account attribution. One failing account doesn't block the rest.
- Read-only — No write tools. Safe for analysis, reporting, and data extraction without risk of modifying production config.
Install
Not yet published to PyPI — install from source for now:
git clone https://github.com/CTMJSON/ctm-multi-mcp
cd ctm-multi-mcp
python3 -m venv .venv
./.venv/bin/pip install -e .
This installs a ctm-multi-mcp executable at .venv/bin/ctm-multi-mcp inside
the cloned folder — use that full path when connecting a Claude client below.
Once published, install will be:
pip install ctm-multi-mcp
# or, with no install, via uv:
uvx ctm-multi-mcp
Configuration
| Variable | Required | Description |
|---|---|---|
CTM_BASIC_AUTH |
yes | Base64-encoded access_key:secret_key for your parent/reseller account. Get the keys from CTM → Settings → Account Settings → API, or generate the encoded value with echo -n "access_key:secret_key" | base64 |
CTM_BASE_URL |
no | Override API base URL (default: https://api.calltrackingmetrics.com/api/v1) |
CTM_ACCOUNT_IDS |
no | Comma-separated account IDs to limit scope |
CTM_ACCOUNTS_FILE |
no | Path to JSON config file (default: ~/.ctm-accounts.json) |
Filtering sub-accounts
To limit which accounts the server queries, set CTM_ACCOUNT_IDS:
export CTM_ACCOUNT_IDS=12345,67890,11111
Or create ~/.ctm-accounts.json:
{
"account_ids": ["12345", "67890", "11111"],
"labels": {
"12345": "Client A (Enterprise)",
"67890": "Client B (SMB)"
}
}
Labels appear in tool output for easy identification. If neither filter is set, all sub-accounts discoverable by the parent key are included.
Connecting to Claude Desktop
Edit your Claude Desktop config file
(macOS: ~/Library/Application Support/Claude/claude_desktop_config.json;
Windows: %APPDATA%\Claude\claude_desktop_config.json).
If the file doesn't exist yet, create it with a mcpServers block and add
an entry for ctm-multi-mcp. Use the full path to the ctm-multi-mcp executable
from the Install step:
{
"mcpServers": {
"ctm-multi-mcp": {
"command": "/absolute/path/to/ctm-multi-mcp/.venv/bin/ctm-multi-mcp",
"env": {
"CTM_BASIC_AUTH": "<base64 access_key:secret_key>"
}
}
}
}
Restart Claude Desktop for the change to take effect. To limit which accounts
are queried, add "CTM_ACCOUNT_IDS": "12345,67890,11111" to env.
Connecting to Claude Code
claude mcp add ctm-multi-mcp \
--env CTM_BASIC_AUTH="<base64 access_key:secret_key>" \
-- /absolute/path/to/ctm-multi-mcp/.venv/bin/ctm-multi-mcp
With account filtering:
claude mcp add ctm-multi-mcp \
--env CTM_BASIC_AUTH="<base64 access_key:secret_key>" \
--env CTM_ACCOUNT_IDS=12345,67890,11111 \
-- /absolute/path/to/ctm-multi-mcp/.venv/bin/ctm-multi-mcp
Once published to PyPI, -- uvx ctm-multi-mcp can replace the explicit path
in both of the examples above.
Connecting to Codex
Edit ~/.codex/config.toml:
[mcp_servers.ctm-multi-mcp]
command = "/absolute/path/to/ctm-multi-mcp/.venv/bin/ctm-multi-mcp"
[mcp_servers.ctm-multi-mcp.env]
CTM_BASIC_AUTH = "<base64 access_key:secret_key>"
CTM_ACCOUNT_IDS = "12345,67890"
Claude.ai (web)
Not currently supported — claude.ai's connectors require a hosted/remote MCP server (reachable over HTTP), while ctm-multi-mcp runs as a local stdio process. Use Claude Desktop or Claude Code instead until a hosted version exists.
Example Prompts
Once connected, you can drive the server in plain English:
- "Give me a call summary for all accounts for the last 7 days."
- "Show me call dispositions across all accounts — what percentage hang up vs answer?"
- "Break down call volume by source across all accounts for August."
- "Which Google/Bing campaigns drove calls for account 12345 this week, and how many were answered?"
- "Export every answered call over 60 seconds for account 12345 yesterday, with name, city, and state."
- "Which accounts have voice bots configured and what are their names?"
- "List all tracking numbers across all accounts with their routing destination."
- "Show me the call logs for account 12345 and 67890 from yesterday."
- "Which accounts have the most sources configured?"
- "Pull all contacts from every account that have 'test' in the name."
Tool Surface
All tools are read-only and operate across sub-accounts:
| Category | Tools |
|---|---|
| Accounts | list_sub_accounts, account_summary |
| Calls | search_calls, get_call, call_export |
| Numbers | list_numbers, get_number |
| Contacts | list_contacts, get_contact |
| Sources | list_sources, get_source |
| Queues | list_queues, get_queue |
| Voice Bots | list_voice_bots, get_voice_bot |
| Voice Menus | list_voice_menus, get_voice_menu |
| Schedules | list_schedules, get_schedule |
| Geo Routes | list_geo_routes, get_geo_route |
| Users | list_users, get_user |
| Tags | list_tags |
| Reports | call_summary_report, call_disposition_report, call_crosstab, reports_series, reports_series_dimensions, account_ledger, account_summary |
| Texts | list_texts |
Every tool accepts an optional account_ids parameter (comma-separated string)
to limit which accounts are queried. Omit to query all discovered accounts.
Working with large result sets
CTM's API caps per_page at roughly 100 records, so a naive integration
either misses data past the first page or has to loop pages by hand. This
server handles that for you in two ways:
- List tools (
list_contacts,list_numbers,list_sources,list_queues,list_users,list_voice_bots,list_voice_menus,list_schedules,list_geo_routes,list_tags,list_texts, and the calls-specificcall_disposition_report/call_crosstab/call_export) page through the CTM API automatically up to amax_items(ormax_calls) cap — default 500 for most, higher for calls — and flag aWARNINGin the output if any account had more records than the cap, so truncation is never silent. Raise the cap rather than paging manually. - Detail tools (
get_call,get_contact,get_number,get_source,get_queue,get_voice_bot,get_voice_menu,get_schedule,get_geo_route,get_user) return the full record by default — which for calls in particular can be 100+ fields including transcripts and touchpoint history — but accept an optionalfieldsparameter (a comma-separated list of top-level keys) to return just what's needed. Cheaper for anything that checks the same few fields across many records.
For call-level analysis specifically, reach for call_disposition_report
(aggregate stats), call_crosstab (two-dimension breakdown, e.g. source ×
status or campaign × status), or call_export (row-level CSV) rather than
paging through search_calls — see each tool's docstring for details on
why (in short: search_calls filters and totals only reflect a single
page, and CTM's calls-list payload already includes fields like name,
city, and state, so there's no need to loop get_call afterward to fill
those in).
call_crosstab/call_export can also group or filter by real ad-platform
attribution — campaign, campaign_id, keyword, ad_group, medium,
paid_source — pulled from CTM's per-call "paid" object, not just CTM's
own tracking-source label.
Architecture
ctm-multi-mcp
├── server.py FastMCP server, credential bootstrap
├── config.py Account filtering config from env/file
├── ctm_client.py Standalone CTM HTTP client (auth, paginate, paginate_capped, rate-limiting)
├── accounts.py Sub-account discovery, resolution, status filtering
├── client.py MultiAccountClient — concurrent operations across accounts
└── tools/
├── accounts.py list_sub_accounts
├── calls.py search_calls, get_call
├── contacts.py list_contacts, get_contact
├── geo_routes.py list_geo_routes, get_geo_route
├── numbers.py list_numbers, get_number
├── queues.py list_queues, get_queue
├── reports.py call_summary_report, call_disposition_report,
│ call_crosstab, call_export, reports_series,
│ reports_series_dimensions, account_ledger,
│ account_summary
├── schedules.py list_schedules, get_schedule
├── sources.py list_sources, get_source
├── tags.py list_tags
├── texts.py list_texts
├── users.py list_users, get_user
├── voice_bots.py list_voice_bots, get_voice_bot
└── voice_menus.py list_voice_menus, get_voice_menu
Support
email: jason.smith@ctm.com
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.