Chipp MCPs (pick only what you need)
The Chipp MCP is split into small MCPs, one per area. Connect core plus only the ones you need, so your AI assistant stays fast and cheap.
Every tool an MCP server lists is sent to your AI model with every message. The full Chipp MCP has hundreds of tools, so connecting all of it costs tens of thousands of input tokens per message before you type a word.
So Chipp is split into small MCPs, one per area. Each has its own URL, lists only its own tools, and uses the same Chipp sign-in and permissions as the full server.
Recommended: connect core (about 2,600 tokens), then add only the MCPs for the work you do. You can add or remove one at any time.
The MCPs
| ID | Name | What it covers | URL | Tokens per message |
|---|---|---|---|---|
core | Core | Who you are, workspaces, finding agents, Chipp docs, and the list of the other Chipp MCPs. | https://mcp.chipp.ai/mcp/core | about 2,600 |
apps | Agents | Create, publish, version, chat with and move agents; system prompt, model, suggestions, variables and agent links. | https://mcp.chipp.ai/mcp/apps | about 8,300 |
app-settings | Agent settings | The full agent settings writers (update_app, bulk_update_app_settings), branding, auth, access lists, tool overrides and workspace defaults. | https://mcp.chipp.ai/mcp/app-settings | about 14,300 |
knowledge | Knowledge | Knowledge sources (URLs, sitemaps, text, documents), crawls, and evidence search over an agent’s knowledge base. | https://mcp.chipp.ai/mcp/knowledge | about 4,200 |
actions | Actions and components | Custom actions, action collections, custom components, and database connections. | https://mcp.chipp.ai/mcp/actions | about 11,100 |
integrations | Integrations | Hosted integrations and their catalog, and MCP servers connected to an agent. | https://mcp.chipp.ai/mcp/integrations | about 2,400 |
analytics | Analytics and evals | Conversations, messages, usage and engagement analytics, end users and their memories, and evals. | https://mcp.chipp.ai/mcp/analytics | about 6,200 |
tags | Tags | Conversation tags, their training examples and utterances, and tag instances. | https://mcp.chipp.ai/mcp/tags | about 4,600 |
channels | Channels | Slack, email, WhatsApp, GitHub, Discord and voice deployments, and heartbeat schedules. | https://mcp.chipp.ai/mcp/channels | about 6,900 |
billing | Billing and monetization | The subscription, credits, usage and capabilities, and consumer packages an agent sells. | https://mcp.chipp.ai/mcp/billing | about 1,700 |
newsletters | Newsletters | Newsletter topics, audiences, contacts, campaigns and sends. | https://mcp.chipp.ai/mcp/newsletters | about 4,200 |
projects | Software factory projects | Software factory projects: list, create and scope them, and the factory tools in no other sub-MCP. | https://mcp.chipp.ai/mcp/projects | about 26,100 |
tickets | Tickets (software factory) | Software factory tickets: list, read, file, reply, retry, cancel, proposals, and child projects. | https://mcp.chipp.ai/mcp/tickets | about 23,500 |
releases | Releases (software factory) | Software factory deploys and releases: history, the deploy gate, Deploy now, rollbacks, recalls and the release digest. | https://mcp.chipp.ai/mcp/releases | about 7,500 |
observability | Observability (software factory) | A deployed project’s database (read), logs, errors, analytics, CI status, custom actions, credentials and cost. | https://mcp.chipp.ai/mcp/observability | about 11,400 |
data-ops | Data operations (software factory) | Changes to a deployed project’s live data: data, script and procedure operations, bulk ingest, and file exports. | https://mcp.chipp.ai/mcp/data-ops | about 11,000 |
packs | Packs (software factory) | The feature pack and recipe catalog, pack installs, checkout, and the project blueprint. | https://mcp.chipp.ai/mcp/packs | about 4,000 |
memory | Memory and schedules (software factory) | A project’s memory, circles, and scheduled tasks. | https://mcp.chipp.ai/mcp/memory | about 5,800 |
legacy-code | Legacy code (software factory) | The code of an app being rebuilt: the imported GitHub repo and the legacy codebase. | https://mcp.chipp.ai/mcp/legacy-code | about 920 |
| (all) | Everything | Every tool in one MCP. | https://mcp.chipp.ai/mcp | about 156,900 |
Token sizes are approximate: they are the tool definitions your model reads on each message, for an account with every permission. The software factory MCPs list tools only when your organization has connected the software factory.
The core MCP has a list_chipp_mcps tool, so you can also ask your assistant “which Chipp MCP do I need for this?”.
Which ones should I add?
| You want to… | Add |
|---|---|
| Find agents and read Chipp docs | core |
| Create agents, edit prompts and models | core, apps |
| Change settings, branding, auth or access | core, app-settings |
| Manage knowledge sources | core, knowledge |
| Read conversations and analytics | core, analytics |
| File and follow software factory tickets | core, tickets |
| Ship, roll back or check deploys | core, releases |
Add one to your client
The steps are the same as for the full server; only the URL changes. In Chipp, Settings > Developer (the MCP connection card) has a picker that copies each URL.
- Claude app / Claude Cowork: add a custom connector with the MCP’s URL, for example
https://mcp.chipp.ai/mcp/core. Add one connector per MCP. See Connect from Claude App or Claude Cowork. - Claude Code:
bash
claude mcp add --transport http chipp-core https://mcp.chipp.ai/mcp/core claude mcp add --transport http chipp-tickets https://mcp.chipp.ai/mcp/tickets - ChatGPT (connectors in developer mode): create a connector with the MCP’s URL and sign in with Chipp.
- Cursor and other clients: use the MCP’s URL as the server URL. See Setup & Installation.
Each MCP asks you to sign in the first time (the same Chipp account and the same approval screen). An API key works on every MCP URL.
The full server
https://mcp.chipp.ai/mcp (also https://build.chipp.ai/mcp) still lists every tool, and existing connections keep working. Use it only if you want everything in one place and accept about 156,900 input tokens per message.