charter-packs-mcp
Provides tools for interacting with GitHub's API, enabling repository, issue, pull request, and workflow management as part of engineering workflows.
Provides tools for interacting with Gmail's API, enabling email reading, sending, and management for support and communication workflows.
Provides tools for interacting with Linear's API, enabling issue tracking, project management, and team collaboration as part of engineering and product workflows.
Provides tools for interacting with Notion's API, enabling page, database, and content management for documentation and knowledge workflows.
Provides tools for interacting with Shopify's API, enabling e-commerce operations such as product, order, and inventory management for commerce workflows.
Provides tools for interacting with Slack's API, enabling messaging, channel management, and notifications for team collaboration workflows.
Provides tools for interacting with Stripe's API, enabling payment processing, subscription management, and billing operations for finance workflows.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
charter-packs-mcp
Charter packs, served over MCP: the runtime behind the R28
MCP servers, with their cross-app workflows and a login that keeps your own keys in your
operating system's keychain.
Looking for the SDK? This is built on Charter, which
turns an API endpoint into a typed tool any agent can call. Start there: pip install charter-ai.
A Charter pack is one API's tools. A family is the tools one kind of work calls across APIs: engineering is GitHub, Linear and Slack together, a support inbox is Gmail, Stripe and Linear. Each family carries the workflows that justify its tools, and its server offers those workflows as prompts (slash commands in Claude Code) beside exactly the tools they name.
Nothing here is part of the Charter SDK. This package is built on it: every tool is a pack tool, and the server is Charter's MCP adapter.
Nothing here is on PyPI: each family installs from a wheel on its own GitHub release, and depends on this package by the wheel on this repository's release. Only Charter comes from PyPI, and installing needs no git.
python -m charter_packs_mcp engineering # serve one family over stdio
python -m charter_packs_mcp support,finance # two families, one server
python -m charter_packs_mcp engineering login # connect the apps it uses
python -m charter_packs_mcp engineering status # what is connectedFamily | Workflows | Published as |
engineering | 25 |
|
product | 20 |
|
sales | 14 |
|
marketing | 13 |
|
support | 12 |
|
finance | 14 |
|
commerce | 12 |
|
ops | 20 |
|
research | 10 |
|
people | 8 |
|
agency | 8 |
|
Connecting apps
Each server tells the agent at startup which apps are connected and how to connect the rest. Every credential is the user's own: nothing here is registered with any service, and there is no server of ours in the path.
Keys (GitHub, Linear, Slack, Stripe, Notion, Firecrawl, Tavily, Granola, Shopify) come from the client's config, or from
login, which asks for them at a hidden prompt.Google signs in through the user's own OAuth client, with PKCE and a loopback redirect, because a shared one would need Google's verification, and for Gmail and Drive a paid security assessment.
Whatever login or the connect tool stores goes to the OS keychain, after one read-only
call to the app's own API confirms it works. A variable set in the client's config always
wins over the keychain.
Related MCP server: AutoTool
Layout
src/charter_packs_mcp/
catalogue.py the 156 workflows: title, tools in order, why, cadence
__init__.py families, the MCP entry point, login / status / logout
apps.py each app: its key, the call that checks it, Google's sign-in server
connections.py where each app's credential came from, handed to the packs
signin.py `login`, Google's sign-in, the `connect` tool, the `setup` prompt
keychain.py the OS keychain, or an owner-only file without one
scripts/generate.py writes ../charter-mcp-families: one publishable package per familyDevelopment
uv sync
uv run pytest
uv run ruff check src tests scripts
uv run pyright
uv run python scripts/generate.pyEvery test runs offline against respx. To work against a Charter checkout rather than the
release, uv pip install -e ../charter after uv sync. Do not put that path in
[tool.uv.sources]: uv honours it when the package is installed from GitHub, where
../charter does not exist, and every install fails.
License
Apache 2.0. See LICENSE.
This server cannot be deployed
Maintenance
Related MCP Connectors
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Zero-setup MCP gateway securely connecting AI to your tools with authentication and workflows
Toolstile is a managed MCP gateway that connects remote MCP servers behind one endpoint, with per-agent permissions, shared accounts and activity logs. Connect at https://mcp.toolstile.com/mcp over Streamable HTTP with OAuth; sign in at https://toolstile.com, select a workspace and connect your upstream servers. Available tools, resources and prompts depend on that workspace and the agent's granted access.
Discover, plan and run tools across your connected business systems through one OAuth MCP endpoint
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables an agent to search approved MCP integrations and expose only the tools a given task needs, signing the user in once via browser-based OAuth and enforcing scopes, effects and budgets in code. It also acts as a single gateway connection so existing hosts can reach many servers with per-tenant isolation and audit events.MIT
- AlicenseAqualityAmaintenanceLets agents create reusable MCP tools, verify and mount them at runtime, and request credentials by name with user consent. Runs locally and caches verified tool servers for reuse.2MIT
- AlicenseNot gradedqualityAmaintenanceConnects AI clients such as Claude Code, Cursor, Codex, ChatGPT, Hermes, and OpenClaw to 2,000+ AI models and data APIs through a local MCP bridge, letting an agent discover, inspect, price, and run tools after a single browser-approved OAuth sign-in. Backed by a read-only catalog, run history, and rollback so calls can be audited and reverted without managing individual API keys.556,724 npmMIT
- AlicenseNot gradedqualityCmaintenanceEnables MCP clients to connect once to self-hosted, user-owned upstream accounts and then discover and invoke granted tools through a single authenticated endpoint. Each client receives its own key with deny-by-default tool grants, while sensitive calls require durable human approval and credentials stay encrypted and never forwarded upstream.Apache 2.0