get_workflow_blocks

get_workflow_blocks is where every workflow build starts. A workflow is an OnFire automation: a trigger (when it runs), then a chain of steps that each take the rows the step before it produced. This tool lists the blocks those steps are built from, and — when you name specific ones — describes each in full.

Workflows are the write side of the MCP: an agent can design one, save it as a draft, and run it. The six workflow tools form a lifecycle:

1. get_workflow_blocks — discover the blocks (this tool).
2. get_workflow_options — the valid values for a block’s settings and inputs.
3. build_workflow — assemble and validate the steps, with a dry-run preview.
4. save_workflow — save them as a draft.
5. run_workflow — run it (a dry run by default).
6. manage_workflows — list, inspect, edit, pause or archive saved workflows.

Tip

Before your first build, read the full playbook: load_onfire_guide("workflows"). It covers step syntax, references between steps, branching, and the governance rules the builder enforces.

Block categories

Category	What it does
Triggers	When the workflow runs — a schedule, a webhook, new CRM records, Signals.
Sources	What records it reads — CRM records, a list you supply, OnFire data.
Enrichments	Add facts to a row — firmographics, a matched company or person.
Transforms	Reshape the rows — SQL steps, derived columns, filters.
Logic	Branch, merge and stop — if, switch, and named stops.
Actions	Write the result out — CRM writes, CSV exports, sequence enrollment.

Two modes

List every block (no types)

Returns each block in one line — its type, display name, category, a short summary, its ports (only where it has more than one way in or a side other than its main row output), and whether your tenant may use it. It also returns the tenant context you build against: the connected CRM and its objects, enabled features, and the variables a step may reference.

Read this once to choose your blocks. It carries no settings, inputs or outputs — full details come only from the second mode.

Describe blocks in full (with types)

Pass the block types you chose — e.g. ["enrich.match_company", "action.csv_export"] — to get each one in full:

• what every parameter and input means, their defaults, and which are required;
• the AnyOf rules between inputs;
• the output fields each block produces;
• the nested parameter_schema for every setting;
• the guides the block’s description points to, and worked examples that use it;
• for a CRM write, your tenant’s CRM field mapping;
• your tenant’s variables.

Read the full detail for every block before you write its step.

Binding inputs to fields

Each input has a tag — the kind of value it accepts (a company website, a person’s LinkedIn URL, and so on). Bind an input only to a field with the same tag. An input tagged text (or untagged) accepts any field.

The values a setting or input can take come from get_workflow_options — never write a CRM object, field, or data name from memory, because an unknown one fails on save. get_workflow_blocks tells you which places need looking up (pick_from_lists gathers them per block); get_workflow_options returns the actual values.

Inputs

Parameter	Type	Required	Description
types	string[]	No	Block types to describe in full. Omit to list every block in one line.
target_tenant_id	string	No	Super-tenant administrators only — act for another tenant. Blocked for everyone else.

Notes

• Skip any block whose available flag is false; unavailable_because says why it isn’t enabled for your tenant.
• The block surface is tenant-scoped: the CRM objects, fields, features and variables you see are your tenant’s.