agent setup
Set up or troubleshoot a Combined remote MCP connection for the user's chosen AI client, preserving the account scope, OAuth flow, and existing source grants. Use when adding Combined to an agent or fixing connection discovery.
THE COMBINED FIELD GUIDE / AGENT SKILLS
Give your agent a working playbook.
Six reusable skills for asking better questions of your business data. Discover the right records, calculate across apps, and return an answer you can check.
Start with your business question. These skills give your agent the steps for finding, joining and querying the sources you have connected to Combined.
Use them with a client that supports agent skills and a Combined MCP connection. Explore Combined, find your source, or follow the Claude Code setup guide.
Install the skills
npx skills add https://www.trycombined.comThe skills CLI lets you select the skills and agents you want to install them for. You can also download each skill below and use your client's normal skill installation flow. Installing a skill does not connect an app or change your source grants.
Install in Claude Code
Use Claude Code's native plugin manager to install the same six playbooks from Combined's public agent kit. Run these commands inside Claude Code:
/plugin marketplace add sqrr-ai/combined-agent-kit
/plugin install combined-business-data@combinedStart a new Claude Code session, then run /combined-business-data:combined-agent-setup for the workspace connection steps. Choose this route or the skills CLI above to keep one copy of the playbooks. Your source connections and grants determine which business records the agent can use.
Inspect the plugin and runnable examples or read the Claude Code marketplace instructions.
business data
Answer a business question across sources already connected to Combined by discovering granted datasets, choosing SQL or text retrieval, and reporting source freshness and coverage. Use for cross-app analysis in a Combined workspace.
customer context
Build a source-backed customer brief from the user's granted Combined CRM, billing, support, and conversation datasets. Use for customer context, account review, or investigating a named company's activity.
hubspot stripe
Compare HubSpot open pipeline with Stripe paid invoices using Combined's granted datasets, explicit customer mapping, separate aggregation, and currency-aware time windows. Use for CRM and billing analysis without double-counting.
readonly sql
Write, troubleshoot, and verify bounded read-only DuckDB SQL through Combined's MCP query_sql tool. Use for granted-dataset joins, aggregates, filters, pagination, truncation, or query-policy errors.
revenue analysis
Analyze billing and CRM metrics from Combined while separating pipeline, invoice collections, recurring revenue, and missing data. Use when a revenue question needs explicit definitions, source reconciliation, and checkable SQL.
Built around your actual data
Each playbook discovers the datasets and fields available in your workspace, checks source freshness, and keeps customer mappings and metric definitions explicit. The HubSpot–Stripe package includes a runnable synthetic fixture for spotting duplicated totals in joins.
Read the HubSpot–Stripe tutorial and Combined's MCP tools reference. The skill discovery index includes download hashes for tooling.
Try the interactive CRM-and-billing example to see how record counts change a join's totals, then download the matching SQL.
Choose your apps in the free MCP planner to get a connection plan, relevant examples and an assignment tailored to your analysis.
Check an answer against its evidence in the free verification lab. Inspect wrong-account, stale-source and unsupported-number examples, then download the checks for your own evaluation.