Skip to main content Event Sunset cruise at Dreamforce 2026
Changelog / 2026 18 updates

What's new

The latest Conversion releases and fixes.

The Conversion MCP server

Conversion now ships a Model Context Protocol server. Connect it to Claude, to an IDE, or to any MCP client, and that client can read and write your marketing data directly.

The point is not another chat surface. It removes the gap between the tool your team already works in and the platform the data lives in, so a question about why a campaign underperformed gets answered from live records rather than a stale export.

What the server exposes

Contacts, companies, opportunities, campaigns, custom objects, audiences, emails, forms, and workflows are all addressable. Reads and writes both work, so a client can go from analyzing a segment to building the workflow that acts on it in one session.

Getting connected

The server is available to every workspace and authenticates with a workspace API key. The rate limits that apply to the public API apply here too. Visit the docs to set it up.

AI Chat is production ready

AI Chat runs against your real workspace now. It knows your objects, fields, campaigns, and audiences without being told, so a question like which segments have not been emailed in thirty days does not start with explaining your schema.

Business context

Context is loaded per workspace and held across a conversation. Chat keeps track of your data model through a long exchange rather than losing it partway.

Thinking mode

Longer questions return a visible reasoning trace before the answer. This matters most for trust: when chat reports that a segment holds four thousand contacts, you can see how it got there instead of re-deriving it.

Statements on analytics

Segment evaluation moved onto the analytics engine. Audience counts on large workspaces return in seconds.

Improvements

  • Audiences An email link clicked statement can specify which link was clicked, not only that some link was.
  • Automations Workflow triggers can filter on event content, and that context carries to later steps in the run.
  • Email Data fields in statement events and utm_content are both exposed to Liquid.
  • Forms A toggle controls whether a form creates a new contact for a unique email or deduplicates against an existing one.
  • Admin Permissions separate members from administrators, so someone can build without being able to reconfigure the Salesforce connection.

Custom events and API-triggered workflows

You can now send events into Conversion from your product, your billing system, or a script, and trigger a workflow on them. The same events are available to segmentation and to Liquid, so a signal that starts a run can also personalize what the run sends.

Custom events

Events arrive through a batch upsert endpoint and appear as engagements on the contact timeline. Payloads support JSON dot notation, so a nested field in the event body can be referenced directly in a statement.

API trigger workflows

A workflow can start from an API call rather than only from a contact-side event. The trigger context is exposed to Liquid inside the workflow, which makes the data that started the run available to every step in it.

CSV Uploads v2

CSV import was rebuilt. Batches run in parallel, the job sends heartbeats so a long import no longer looks stalled, and column headers are detected rather than assumed.

Improvements

  • Automations New workflow nodes for creating a task, assigning to a queue, and linking to a campaign.
  • Automations A validation endpoint catches broken configuration before a workflow goes live.
  • Data Model Contact variable change events fire on update, so a workflow can react to a field changing instead of polling for it.
  • Email A recipient allowlist restricts who a workspace can send to, which is what teams want while testing.
  • Admin OAuth errors for Salesforce and Slack surface what went wrong instead of failing quietly.
  • Admin Default views are configurable per workspace.

Public API v2 and warehouse query previews

Warehouse queries can now be previewed before they run. You write the query, see the rows it returns, and then decide whether to sync, which turns a slow feedback loop into an immediate one.

Warehouse query preview

Results come back in the builder. Warehouse pages are split into work units and fetches are cached, so a preview against a large table returns rather than timing out.

Public API v2

Contacts move to v2, with cursor pagination and a cleaner response shape. An endpoint for external field schemas lets an integration discover what fields exist instead of hard-coding them.

Rate limits are tracked per workspace and returned on every response.

Form identity resolution

Form submissions resolve to an existing contact more reliably. The same person filling out two different forms produces one contact rather than two.

Improvements

  • Campaigns Campaign members can be exported, and campaign member CSV upload supports Liquid fields.
  • Email Preview renders upversioned content, so what you see matches what sends.
  • Data Model Bot detection tightened from a ten second click window to five, which cuts false positives from scanners without discarding real early clicks.
  • Data Model Contact engagement moved to v3, and repeated engagements of the same type are kept rather than collapsed.
  • CRM Sync Salesforce org API limits are cached and exposed through an endpoint, so you can see how close a sync is to your org’s daily quota.

Fixes

  • Automations Retries are slower and more numerous, which resolves runs killed by a transient Salesforce error.
  • Data Model Event ordering uses the message timestamp rather than processing time, so events stay in sequence under load.

Snowflake and BigQuery sync

You can now connect Snowflake or BigQuery and sync warehouse data onto contacts and companies. The pipeline runs end to end, and each stage reports separately, so a slow sync tells you which stage is slow.

Warehouse integrations

Snowflake and BigQuery authenticate through the warehouse API. A sync is its own resource with its own configuration and run history, and the extract, transform, and load stages are observable independently.

A deeper object model

Object definitions and the relationships between them moved onto a new schema this month. The practical effect is deeper relationship modelling between custom objects and the built-in ones, and faster reads when a segment traverses several of them.

Salesforce syncer v3

The syncer was rewritten. Validation now happens before records are pushed rather than after they fail, and OAuth errors come back with the provider’s error code instead of a generic redirect.

Improvements

  • Email A tracking domain can be set up before the base domain is verified, which unblocks the slowest step in onboarding.
  • Email Email events flow through the event pipeline, so engagement reaches segmentation faster.
  • Email A validate endpoint for Liquid catches template errors before a send rather than during one.
  • Audiences Stored statements convert to SOQL, which is what makes Salesforce-side evaluation possible.
  • Audiences Audiences sync on a schedule rather than only on demand.
  • Integrations LinkedIn ad audiences are available through external endpoints.
  • Data Model Bot activity is included in email events so it can be filtered rather than silently dropped.

Fixes

  • Audiences Null values sort last consistently.
  • Forms Submissions from untracked forms are marked not found instead of returning an error.
  • CRM Sync Empty values are allowed where Salesforce permits them.

Campaign member sync, and opportunities

Campaign members now sync in both directions with Salesforce. A status change made in either system appears in the other on the next sync, so nobody has to export a list to reconcile the two.

Bidirectional campaign member sync

Creates, updates, and deletes all move through the syncer. Campaign member fields are available in Liquid, so an email can personalize on where someone sits in a campaign.

Linking an existing Salesforce campaign to a Conversion campaign preserves the membership already on it. You no longer rebuild it from scratch.

Opportunities across the data model

Opportunities and opportunity contact roles are real objects now, with their own schema, default variables, and relationship endpoints to contacts and companies. Segmentation and reporting can reference pipeline directly rather than through a workaround.

Deletions propagate. An opportunity removed in Salesforce is removed in Conversion.

Sync logs

Every sync run now writes a log you can open. When a record fails, the reason is on the record instead of somewhere in a batch summary.

Improvements

  • Email One-click unsubscribe is on for every workspace, and the flag that gated it is gone.
  • Email Operational email can send without an unsubscribe link, which is what compliance teams expect for transactional messages.
  • Data Model Communication limits apply per recipient email rather than per contact record.
  • Campaigns Duplicating a campaign brings its assets with it.
  • Admin Sync settings respect workspace permissions.
  • Audiences Statement evaluation moved off the legacy engine, and the old path is deprecated.

Fixes

  • Campaigns Deleting a campaign member uses the member id rather than matching on contact, so the wrong record is no longer removed when a contact appears twice.
  • CRM Sync Contacts missing fields Salesforce requires are filtered out before the create attempt instead of failing the batch.
  • CRM Sync All fields are written to the field map when unmarshalling contacts, leads, and accounts.
Newer Page 2 of 2