Platform
The answer where the question is, not in a fourteenth tab.
A metric you have to go and look at is a metric you look at on Fridays. Pacia exposes its analytics over the Model Context Protocol, so an AI-enabled editor can answer "what is at risk in this sprint" without you leaving the file you are in — and every tool calls the same service the equivalent page calls, so the two answers cannot disagree.
- VS Code & JetBrains
- Extension and plugin
- Claude Code & Cursor
- Over MCP
- Slack & Microsoft Teams
- PR queue and alerts
- CI pipelines
- @pacia/cli coverage reporting
- Organisation scoping
- From your token, never a tool argument
Where Pacia reaches
MCP server
A stateless Streamable HTTP endpoint exposing the delivery analytics as MCP tools. Works with any MCP-capable client, including Claude Code and Cursor.
VS Code extension
Delivery context and your own queue inside the editor.
JetBrains plugin
The same, for IntelliJ-family IDEs.
Slack app
Pull request queue, lifecycle alerts and a configurable pickup warning, per team channel.
Microsoft Teams app
One centrally operated Pacia bot. Your admin uploads an app package — there are no Azure resources for you to create and no client secrets to supply.
Pacia CLI
The public @pacia/cli npm package reports test coverage from your CI pipeline with an authenticated key.
The editor and the web page share one answer
MCP is another route into the product services, not a second analytics implementation. Identity, organisation scope and the calculation remain in the same place.
One calculation, wherever the question is asked
- 1Editor or chatVS Code, JetBrains, Claude Code, Cursor
- 2Authenticated MCP callIdentity and membership checked live
- 3Product serviceThe same organisation-scoped service the page calls
- 4Matching answerEditor and web cannot invent separate totals
Organisation is resolved from the caller’s token, never from a model argument. MCP tools delegate to product services rather than reimplementing analytics.
How access is kept safe
Organisation never comes from the model
An organisation identifier is never a tool argument. It is taken from the caller’s token. A language model will cheerfully invent a plausible organisation id, and this is exactly the shape of mistake that causes a cross-tenant leak.
Two ways in, one row to revoke
A personal token minted at Settings → MCP access, or OAuth 2.1 with PKCE where the editor opens a browser and you sign in normally. Both end as a single access-token row, so revocation has one implementation rather than two.
Re-checked live on every request
The MCP endpoint cannot run an interactive login or an MFA challenge — JSON-RPC has no channel for one. So MFA is satisfied at issuance, and active status, organisation membership and the MFA requirement are re-read from the database on every single call.
Tools delegate, they never compute
Each MCP tool calls the same service the equivalent page calls. That is what stops a figure in your editor drifting from the figure on screen — there is only one implementation of each calculation.
Common questions
What is MCP?
The Model Context Protocol — an open standard that lets an AI assistant call external tools. Pacia runs a standard MCP server, so any compatible client can query your delivery analytics with your permissions.
Can a model see another organisation’s data?
No. The organisation is resolved from the token on every request and is not something a tool call can specify. Membership and active status are re-verified per call rather than trusted from an earlier check.
Do the editor integrations cost extra?
No. The MCP server, extensions, chat apps and CLI are all part of the standard price.