Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.chartcastr.com/llms.txt

Use this file to discover all available pages before exploring further.

Looking for setup steps? See AI Tools — Setup.
Chartcastr exposes a Model Context Protocol (MCP) server that lets any MCP-compatible AI tool — Claude Code, Cursor, ChatGPT, Codex — read your account data directly. Sources, destinations, connections, and the latest pulse analysis are all available as structured tool calls. This unlocks two distinct patterns.

​
Pattern 1 — Context layer

Your AI assistant knows more when it knows what’s happening in your business. Without MCP, you describe your data situation manually. “Revenue is down 12% this week, conversion has slipped, and we’re seeing lower engagement on the enterprise tier.” You become the bridge — copying from dashboards, summarising in plain text, hoping you captured the nuance. With Chartcastr connected via MCP, your agent can fetch that analysis itself. It can pull the latest pulse — which includes not just the raw numbers but Chartcastr’s contextual analysis of what changed, why it matters in the context of prior trends, and what’s anomalous — and use that as the starting point for whatever you’re asking it to do. This matters for coding tasks, copywriting, strategic planning, customer emails, and anything else where business context shapes the output.

​
Pattern 2 — Action agents

This is where it gets more interesting. Chartcastr runs on a schedule. Every morning (or every hour, or every week — you configure it), it ingests data from your sources, runs AI analysis against your historical context, and produces a pulse: a structured, narrative interpretation of your business right now. Previously, that pulse went to Slack or email. A human read it, decided what to do, and acted. Now you can put an agent in that loop. An action agent connects to Chartcastr via MCP, fetches the latest pulse for one or more connections, interprets the analysis, and decides what to do next:
  • If CAC is trending up and conversion is dropping, trigger a deep-dive analysis script
  • If revenue is significantly above target, generate and schedule a team celebration post
  • If a key metric crosses a threshold, open a GitHub issue or create a Linear ticket
  • If the pulse flags an anomaly, send an escalation to a specific Slack channel with additional context
Chartcastr has already done the hard work: ingesting the data, normalising it, running it against prior periods, applying your business context, and producing a human-readable interpretation. Your agent doesn’t need to recreate that pipeline. It just needs to read the output and decide what to do. 
Your data sources
       ↓
Chartcastr (scheduled analysis, context, AI interpretation)
       ↓
MCP → Your AI agent
       ↓
Actions (Slack, GitHub, Linear, email, scripts, webhooks…)

​
What’s exposed via MCP

ToolWhat it returns
list_sourcesEvery data source with status and connection count
list_destinationsEvery configured destination (Slack workspace, email, etc.)
list_connectionsEvery active chart delivery — source, destination, cadence
get_latest_pulseThe full AI narrative analysis from the most recent delivery
verify_connectionLive confirmation that a specific connection is active
open_chartcastrDeep link to open any part of the dashboard
get_latest_pulse is the richest — it returns Chartcastr’s full narrative analysis, not just raw numbers. This is what makes the action agent pattern work: the agent gets interpreted context, not a spreadsheet.

​
The bigger picture

Most AI tools are powerful but context-blind. They know how to reason, but they don’t know what’s happening in your business right now. Chartcastr is the layer that provides that context — continuously, on schedule, with full access to your data stack. The MCP integration is the bridge between Chartcastr’s analysis and the agents that can act on it. Used as a simple context layer, it makes every AI conversation more grounded. Used to power action agents, it closes the loop from data → insight → action entirely. Ready to set it up? Three-step install →