Open app
Documentation

Setting up Grunda

This guide takes you from an empty workspace to a governed CRM: connected, synced, findings reviewed, fixes approved, and the governed API serving clean answers to your tools. Most teams complete it in under an hour — the first full sync of a large portal runs in the background while you work through the rest.

Read-only by default. Connecting a CRM never gives Grunda standing write access to your data. Every change is proposed as a finding and executed only after someone on your team approves it.

1. Get access

Grunda is invite-only. An administrator in your organization invites you by email from Settings → Members; if your company is new to Grunda, book a demo and we'll set up your organization and send the first invite.

  • Invite links are single-use and expire after 14 days.
  • Sign in with the exact email address that was invited — invites are bound to it.
  • You can sign in with Google, Microsoft, or an email password, whichever your invite allows.

2. Connect a data source

Go to Settings → Connectors and pick your system. Which connectors appear depends on what's enabled for your organization.

HubSpot

  1. Click Connect HubSpot and choose the portal to connect.
  2. Review the requested scopes and grant access. Grunda asks for read scopes on CRM objects, properties, lists, forms, and workflows — the write scopes are used only when you approve a fix.
  3. You're redirected back to Grunda and the portal appears as connected.

Salesforce, Planhat, Pipedrive

Same flow: Connect, authorize in the provider's own window, land back in Grunda. Planhat connects with an API token instead of OAuth — create one in Planhat under Settings, then paste it into the connect dialog.

CSV upload

No CRM, or a dataset that lives outside it? Upload a CSV of companies and Grunda treats it as a portal: same findings, same approval queue, same governed API.

3. Run the first sync

The first sync starts automatically after connecting. Grunda mirrors your objects (companies, contacts, deals), properties, and associations, then computes fill rates and a data health score.

  • Large portals are fine. Syncs are resumable background jobs — portals with hundreds of thousands of records complete across multiple runs, and the mirror is never empty mid-sync.
  • After the first sync, incremental syncs pick up changes continuously, deletion webhooks keep the mirror honest, and a scheduled full sync sweeps up anything missed.
  • Progress is visible on the portal overview; you'll get a notification when the first sync completes.

4. Review findings

The Insights page is the daily surface: every finding ranked by business impact, at two levels.

  • Field level — fill rates against targets, redundant fields describing the same concept, taxonomy drift (one market split across four spellings).
  • Record level — duplicate companies and contacts, records owned by people who left, records whose domains no longer resolve.

Findings you don't care about can be dismissed or snoozed — they stay gone until the underlying data changes materially.

5. Approve fixes

Every proposed change lands in the Queue. Nothing is written back to your CRM until someone approves it there.

  • Merges pick a winner by your policy (most complete, most recent, or custom) and preserve activity history and associations.
  • Field consolidations repoint everything that referenced the losing field — workflows, lists, forms, calculated properties — before archiving it.
  • Normalizations rewrite variant values to the canonical one, with the original preserved in provenance.
  • If your CRM's API can't do something (for example, merging certain object types), Grunda tells you exactly what it couldn't do and why — it never half-applies a fix.

6. Define segments & lock taxonomies

Segments turn tribal knowledge into explicit, stored conditions — “Enterprise customer” becomes a filter everyone shares, used to scope findings, fill-rate targets, and alerts. Taxonomies lock a field's allowed values, so “US”, “USA”, and “United States” can never diverge again: new variants show up as violations, ready to normalize.

7. Automation rules

Once you've approved the same kind of fix a few times, promote it to a standing rule under Automations. Rules run after every sync and queue their results for approval — or execute automatically once you trust them. Typical rules: normalize country variants, enrich a blank industry on active customers, flag duplicate pairs above a match threshold.

8. Governed API & MCP

Create an API key in Settings → API keys. Keys are scoped to one portal and are read-only.

Governed REST API
# What fields exist, what they mean, what values are allowed
curl -H "Authorization: Bearer $KEY" https://app.grunda.ai/api/v1/schema

# One company, governed — by record id or domain
curl -H "Authorization: Bearer $KEY" https://app.grunda.ai/api/v1/companies/acme.com

Responses carry provenance (which raw CRM property each value came from, what it superseded, whether a taxonomy rewrote it) and quality (completeness, last modified, open findings), so a caller can tell a solid answer from a guess.

For agents, point any MCP-compatible client (Claude, or your own agent framework) at https://app.grunda.ai/api/mcp with the same key. The MCP server exposes read-only tools over the governed view: search companies, field catalog, top values, quality reports, segments, and taxonomy reports.

9. Slack & Microsoft Teams

Connect a workspace under Settings → Notifications to get findings, sync results, and approval requests where your team already works. In Slack and Teams you can also ask Grunda questions in chat — answers run read-only, mapped to your own Grunda user and permissions. Teams notifications work with just a Workflows webhook URL if you don't want to install the full bot.

10. Self-hosting & data residency

Enterprise customers can run the entire Grunda backend in their own cloud with Docker, or stay on our cloud while the CRM mirror lives in their own Postgres, in their own region. Migrations apply automatically on boot, and the app runs behind your own load balancer with a standard health endpoint. Contact us for the self-hosting guide and an architecture review.

Getting help

Stuck on any step? Head to the support page for how to reach us and what to include — or book time with the team and we'll walk through the setup with you on your own data.