Building Workflows

Overview

Workflows are persistent, versioned automations that can be triggered on a schedule, via API webhook, by email, by application events, or through web forms. Unlike code execution (ephemeral, one-off scripts), workflows are saved, re-runnable, and editable in the Pinkfish web app. For agents that build workflows programmatically, see the Workflow Builder Prompt.

	Code Execution	Workflows
Purpose	One-off scripts, data transformation	Persistent, reusable automations
Lifetime	Ephemeral — runs once and discards	Permanent — saved, versioned, re-runnable
Triggers	None — you call it when you need it	Schedule, API webhook, email, app events, web forms
Visual UI	None	Full node graph in the Pinkfish web app
Multi-step	Single script with `callTool()`	Named nodes connected by edges
State	No persistence between runs	Inputs, outputs, pinned results persist
File I/O	None	`pf.files.readFile()`, `pf.files.writeFile()`
Resource bindings	None	Connections, triggers, agents, collections

Workflow Tools Reference

All workflow tools are on the /pinkfish-sidekick server path. For full parameter schemas, see the pinkfish workflows server reference.

Tool	Description
Discovery
`capabilities_discover`	AI-powered discovery — describe a task in natural language, get matching tools, connections, resources, and skills
`capability_details`	Get full parameter schemas, connection metadata, and usage instructions for specific tools/connections
Workflow Lifecycle
`workflow_create`	Create a new workflow. Returns `automationId` and starter code
`workflow_update`	Update workflow code and/or bind resources (connections, collections)
`workflow_read`	Read workflow structure — nodes, edges, resources, and optionally full code
`workflow_run`	Execute the workflow and return results
`workflow_run_status`	Check status of a running workflow
`workflow_set_inputs`	Set default input values or define an input schema
`workflow_pin`	Pin node outputs to reuse across runs
`workflow_results`	Deep inspection of workflow run outputs — search, read data, get signed URLs
`workflow_edit`	Surgical edits to workflow structure — add/remove nodes, edges, and resources
`workflow_list`	List all workflows accessible to the user
Triggers
`workflow_trigger_schedule`	Create/manage cron-based schedule triggers
`workflow_trigger_api`	Create/manage API webhook triggers
`workflow_trigger_email`	Create/manage email triggers
`workflow_trigger_application`	Create/manage app event triggers
`workflow_trigger_interface`	Create/manage web form triggers
`workflow_trigger_list_all`	List all triggers for a workflow
`workflow_trigger_cleanup`	Remove orphaned or broken triggers
Agents & Sub-Workflows
`workflow_agents`	Create, read, update, list, and invoke AI agents
`workflow_invoke`	Invoke another workflow (sub-workflow) via its webhook trigger URL

How Workflows Work

A workflow is a directed graph of nodes connected by edges, written as JavaScript using the pf SDK.

Node Types

Type	Purpose	MCP Calls?
`trigger`	Entry point (manual, schedule, API webhook, email, app event, form)	No
`mcp-tool`	Calls exactly one MCP tool (can call it multiple times in a loop)	Yes — one tool
`code-block`	Custom JavaScript for data transformation (no MCP calls)	No
`if-else`	Binary branching based on a condition	No
`router`	Multi-way branching (like switch/case)	No
`for-each`	Iterate over an array, execute body nodes per item	No
`merge`	Reconverge after branching	No
`loop`, `while`, `parallel`, `delay`, `sub-workflow`	Advanced control flow	No

Each mcp-tool node calls exactly one MCP tool. If you need to call two different tools, use two separate nodes connected by an edge.

inputSchema Sources

Each field in a node’s inputSchema must declare a source:

Source	Pattern	Example	Description
`"literal"`	`"value"`, `5`, `true`	`"Daily Report"`	Hardcoded constant
`"node"`	`@node_<name>.field`	`@node_fetch_emails.emails`	Output from another node
`"input"`	`@input.field`	`@input.userEmail`	Value from trigger/POST body
`"resource"`	`{{resource.X}}`	`{{resource.emailConn}}`	Bound connection, trigger, or agent

Resource Bindings

Resources declare what external connections, triggers, and agents a workflow uses. They are declared in WORKFLOW_RESOURCES and referenced in node parameters using the {{resource.X}} pattern.

const WORKFLOW_RESOURCES = {
  gmailConn: {
    type: "connection",
    application: "gmail",
    description: "Gmail account for sending emails",
  },
  apiEndpoint: {
    type: "trigger",
    triggerType: "api",
    description: "API webhook endpoint",
  },
  analyst: {
    type: "agent",
    description: "AI agent for analysis",
  },
};

When you call workflow_update with bindings, you map each resource key to a real ID:

{
  "bindings": {
    "gmailConn": "pcid_abc123",
    "apiEndpoint": "trigger_xyz",
    "analyst": "agent_456"
  }
}

The `pf` SDK

The pf SDK is available inside all node functions:

Function	Returns	Description
`await pf.files.writeFile(filename, data)`	`{ fileId, filename, mimeType, size }`	Write output — every node must call this
`await pf.files.readFile(fileId)`	parsed content	Read a file from a previous node
`await pf.files.getFileUrl(fileId)`	signed URL	Get a downloadable URL (~1 hour expiry)
`await pf.mcp.callTool(serverName, toolName, args)`	tool result	Call an MCP tool (only in `mcp-tool` nodes)
`pf.log.info(msg)`	void	Log info message
`pf.log.success(msg)`	void	Log success message
`pf.log.error(msg)`	void	Log error message
`pf.log.warn(msg)`	void	Log warning message

Workflow Code Structure

Every workflow follows this structure:

//---REQUIRED HEADER - DO NOT MODIFY---
import { pf } from "./pf-bootstrap.mjs";
//---END REQUIRED HEADER---

// 1. Declare resources (connections, triggers, agents)
const WORKFLOW_RESOURCES = {
  // resource declarations...
};

// 2. Define nodes (the steps in your workflow)
const WORKFLOW_NODES = [
  { id: "trigger_1", name: "Start", type: "trigger", triggerType: "manual" },
  // ... more nodes
];

// 3. Define edges (execution order)
const WORKFLOW_EDGES = [
  { source: "trigger_1", target: "node_step_one" },
  // ... more edges
];

// 4. Write node functions
async function node_step_one(params) {
  // Your logic here...
  await pf.files.writeFile("node_step_one_output.json", result);
  return result;
}

// 5. Register functions in global scope
global.node_step_one = node_step_one;

//---REQUIRED FOOTER - DO NOT MODIFY---
await pf.run(WORKFLOW_NODES, WORKFLOW_EDGES);
//---END REQUIRED FOOTER---

Example: Create and Run a Workflow

Step 1: Discover tools for the task

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "capabilities_discover",
      "arguments": {
        "request": "search the web for AI news and summarize the results"
      }
    },
    "id": 1
  }'

Step 2: Get full schemas

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "capability_details",
      "arguments": {
        "items": ["web-search", "embedded-groq"]
      }
    },
    "id": 1
  }'

Step 3: Create the workflow

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_create",
      "arguments": {
        "name": "Daily News Summary",
        "description": "Search the web for AI news and summarize the results"
      }
    },
    "id": 1
  }'

Response:

{
  "id": "auto_abc123",
  "starterCode": "// ... starter workflow template ..."
}

Step 4: Update with code

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_update",
      "arguments": {
        "automationId": "auto_abc123",
        "name": "Daily News Summary",
        "changeDescription": "Add search and summarize nodes",
        "code": "<FULL WORKFLOW CODE>"
      }
    },
    "id": 1
  }'

Step 5: Run the workflow

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_run",
      "arguments": {
        "automationId": "auto_abc123"
      }
    },
    "id": 1
  }'

Step 6: Inspect the results

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_results",
      "arguments": {
        "automationId": "auto_abc123",
        "operation": "read",
        "filename": "summary.json"
      }
    },
    "id": 1
  }'

Example Script

Create, update, run, and inspect results. Requires curl and jq.

#!/bin/bash
API_KEY="<YOUR_API_KEY>"
ORG_ID="<YOUR_ORG_ID>"
API_URL="https://app-api.app.pinkfish.ai"
MCP_URL="https://mcp.app.pinkfish.ai"

# 1. Get token
PINKFISH_TOKEN=$(curl -s -X POST "${API_URL}/auth/token" \
  -H "X-Api-Key: ${API_KEY}" \
  -H "X-Selected-Org: ${ORG_ID}" \
  -H "Content-Type: application/json" | jq -r '.token')
if [ "$PINKFISH_TOKEN" = "null" ] || [ -z "$PINKFISH_TOKEN" ]; then
  echo "Error: Failed to get token"; exit 1
fi

H="Authorization: Bearer ${PINKFISH_TOKEN}"
HCT="Content-Type: application/json"
HA="Accept: application/json"

# 2. Create workflow
R=$(curl -s -X POST "${MCP_URL}/pinkfish-sidekick" -H "$H" -H "$HCT" -H "$HA" \
  -d '{"jsonrpc":"2.0","method":"tools/call","params":{"name":"workflow_create","arguments":{"name":"Daily News Summary","description":"Search and summarize AI news"}},"id":1}')
AUTO_ID=$(echo "$R" | jq -r '.result.content[0].text' | jq -r '.id // empty')
[ -z "$AUTO_ID" ] && AUTO_ID=$(echo "$R" | jq -r '.result.structuredContent.id // empty')
[ -z "$AUTO_ID" ] && { echo "Error: workflow_create failed"; echo "$R" | jq .; exit 1; }
echo "Created workflow: $AUTO_ID"

# 3. Update with code (search + summarize workflow)
WORKFLOW_CODE='//---REQUIRED HEADER - DO NOT MODIFY---
import { pf } from "./pf-bootstrap.mjs";
//---END REQUIRED HEADER---

const WORKFLOW_RESOURCES = {};
const WORKFLOW_NODES = [
  { id: "trigger_1", name: "Start", type: "trigger", triggerType: "manual" },
  { id: "node_search_news", name: "Search News", type: "mcp-tool", serverName: "web-search", toolName: "search_googlesearch", parameters: { query: "latest AI news 2026" }, inputSchema: { query: { type: "string", source: "literal", value: "latest AI news 2026" } } },
  { id: "node_summarize", name: "Summarize", type: "mcp-tool", serverName: "embedded-groq", toolName: "embedded-groq_generate", parameters: { prompt: "@node_search_news", systemPrompt: "Summarize into 5 bullet points." }, inputSchema: { prompt: { type: "string", source: "node", value: "@node_search_news" }, systemPrompt: { type: "string", source: "literal", value: "Summarize into 5 bullet points." } } }
];
const WORKFLOW_EDGES = [
  { source: "trigger_1", target: "node_search_news" },
  { source: "node_search_news", target: "node_summarize" }
];

async function node_search_news(params) {
  const result = await pf.mcp.callTool("web-search", "search_googlesearch", { query: params.query });
  await pf.files.writeFile("node_search_news_output.json", result);
  return result;
}
async function node_summarize(params) {
  const result = await pf.mcp.callTool("embedded-groq", "embedded-groq_generate", { prompt: JSON.stringify(params.prompt), systemPrompt: params.systemPrompt });
  await pf.files.writeFile("node_summarize_output.json", result);
  return result;
}
global.node_search_news = node_search_news;
global.node_summarize = node_summarize;

//---REQUIRED FOOTER - DO NOT MODIFY---
await pf.run(WORKFLOW_NODES, WORKFLOW_EDGES);
//---END REQUIRED FOOTER---'

R=$(curl -s -X POST "${MCP_URL}/pinkfish-sidekick" -H "$H" -H "$HCT" -H "$HA" \
  -d "{\"jsonrpc\":\"2.0\",\"method\":\"tools/call\",\"params\":{\"name\":\"workflow_update\",\"arguments\":{\"automationId\":\"$AUTO_ID\",\"name\":\"Daily News Summary\",\"changeDescription\":\"Add nodes\",\"code\":$(echo "$WORKFLOW_CODE" | jq -Rs .)}},\"id\":1}")
[ -n "$(echo "$R" | jq -r '.error // empty')" ] && { echo "Error: workflow_update failed"; echo "$R" | jq .; exit 1; }
echo "Updated workflow"

# 4. Run workflow
R=$(curl -s -X POST "${MCP_URL}/pinkfish-sidekick" -H "$H" -H "$HCT" -H "$HA" \
  -d "{\"jsonrpc\":\"2.0\",\"method\":\"tools/call\",\"params\":{\"name\":\"workflow_run\",\"arguments\":{\"automationId\":\"$AUTO_ID\"}},\"id\":1}")
[ -n "$(echo "$R" | jq -r '.error // empty')" ] && { echo "Error: workflow_run failed"; echo "$R" | jq .; exit 1; }
echo "Started workflow run"

# 5. Wait for completion, then read results
sleep 15
R=$(curl -s -X POST "${MCP_URL}/pinkfish-sidekick" -H "$H" -H "$HCT" -H "$HA" \
  -d "{\"jsonrpc\":\"2.0\",\"method\":\"tools/call\",\"params\":{\"name\":\"workflow_results\",\"arguments\":{\"automationId\":\"$AUTO_ID\",\"operation\":\"read\",\"filename\":\"node_summarize_output.json\"}},\"id\":1}")
echo "Summary output:"
echo "$R" | jq '.result.structuredContent // .result.content[0].text'

Complete Code Example: Web Search + Summarize

//---REQUIRED HEADER - DO NOT MODIFY---
import { pf } from "./pf-bootstrap.mjs";
//---END REQUIRED HEADER---

const WORKFLOW_RESOURCES = {};

const WORKFLOW_NODES = [
  {
    id: "trigger_1",
    name: "Start",
    type: "trigger",
    triggerType: "manual",
  },
  {
    id: "node_search_news",
    name: "Search News",
    type: "mcp-tool",
    serverName: "web-search",
    toolName: "search_googlesearch",
    parameters: {
      query: "latest AI news 2026",
    },
    inputSchema: {
      query: {
        type: "string",
        source: "literal",
        value: "latest AI news 2026",
      },
    },
  },
  {
    id: "node_summarize",
    name: "Summarize Results",
    type: "mcp-tool",
    serverName: "embedded-groq",
    toolName: "embedded-groq_generate",
    parameters: {
      prompt: "@node_search_news",
      systemPrompt: "Summarize these search results into 5 bullet points.",
    },
    inputSchema: {
      prompt: { type: "string", source: "node", value: "@node_search_news" },
      systemPrompt: {
        type: "string",
        source: "literal",
        value: "Summarize these search results into 5 bullet points.",
      },
    },
  },
];

const WORKFLOW_EDGES = [
  { source: "trigger_1", target: "node_search_news" },
  { source: "node_search_news", target: "node_summarize" },
];

async function node_search_news(params) {
  pf.log.info("Searching for AI news...");
  const result = await pf.mcp.callTool("web-search", "search_googlesearch", {
    query: params.query,
  });
  await pf.files.writeFile("node_search_news_output.json", result);
  pf.log.success("Search complete");
  return result;
}

async function node_summarize(params) {
  pf.log.info("Summarizing results...");
  const result = await pf.mcp.callTool(
    "embedded-groq",
    "embedded-groq_generate",
    {
      prompt: JSON.stringify(params.prompt),
      systemPrompt: params.systemPrompt,
    },
  );
  await pf.files.writeFile("node_summarize_output.json", result);
  pf.log.success("Summary generated");
  return result;
}

global.node_search_news = node_search_news;
global.node_summarize = node_summarize;

//---REQUIRED FOOTER - DO NOT MODIFY---
await pf.run(WORKFLOW_NODES, WORKFLOW_EDGES);
//---END REQUIRED FOOTER---

Complete Code Example: With Resource Bindings and Triggers

This example shows a workflow with a connection binding (Zendesk), an API trigger, and an agent:

//---REQUIRED HEADER - DO NOT MODIFY---
import { pf } from "./pf-bootstrap.mjs";
//---END REQUIRED HEADER---

const WORKFLOW_RESOURCES = {
  zendesk: {
    type: "connection",
    application: "zendesk",
    description: "Zendesk support account",
  },
  apiEndpoint: {
    type: "trigger",
    triggerType: "api",
    description: "API endpoint — bind via workflow_trigger_api tool",
  },
  ticketAnalyst: {
    type: "agent",
    description:
      "Agent that analyzes support tickets for patterns and insights",
  },
};

const WORKFLOW_NODES = [
  {
    id: "trigger_1",
    name: "API Webhook",
    type: "trigger",
    triggerType: "api",
    inputSchema: {
      ticketLimit: {
        type: "number",
        description: "Number of tickets to fetch",
      },
    },
  },
  {
    id: "node_fetch_tickets",
    name: "Fetch Zendesk Tickets",
    type: "mcp-tool",
    serverName: "zendesk",
    toolName: "zendesk_list_tickets",
    parameters: {
      PCID: "{{resource.zendesk}}",
      per_page: "@trigger_1.ticketLimit",
    },
    inputSchema: {
      PCID: { type: "string", source: "resource" },
      per_page: { type: "number", source: "input" },
    },
  },
  {
    id: "node_format_tickets",
    name: "Format Tickets",
    type: "code-block",
    parameters: {
      tickets: "@node_fetch_tickets.tickets",
    },
    inputSchema: {
      tickets: { type: "array", source: "node" },
    },
  },
  {
    id: "node_analyze_tickets",
    name: "Analyze Tickets",
    type: "code-block",
    parameters: {
      agentId: "{{resource.ticketAnalyst}}",
      message:
        "Analyze these support tickets and identify top issues, urgent count, and trends.",
      formattedTickets: "@node_format_tickets.formattedTickets",
    },
    inputSchema: {
      agentId: { type: "string", source: "resource" },
      message: { type: "string", source: "literal" },
      formattedTickets: { type: "string", source: "node" },
    },
  },
];

const WORKFLOW_EDGES = [
  { source: "trigger_1", target: "node_fetch_tickets" },
  { source: "node_fetch_tickets", target: "node_format_tickets" },
  { source: "node_format_tickets", target: "node_analyze_tickets" },
];

async function node_fetch_tickets(params) {
  pf.log.info("Fetching Zendesk tickets...");
  const result = await pf.mcp.callTool("zendesk", "zendesk_list_tickets", {
    ...params,
    version: "1",
  });
  await pf.files.writeFile("node_fetch_tickets_output.json", result);
  pf.log.success("Fetched " + (result.tickets?.length || 0) + " tickets");
  return result;
}

async function node_format_tickets(params) {
  const { tickets } = params;
  pf.log.info("Formatting " + tickets.length + " tickets...");
  const formattedTickets = tickets
    .map(
      (t) =>
        `[${t.id}] ${t.subject} - Status: ${t.status}, Priority: ${t.priority}`,
    )
    .join("\n");
  const output = { formattedTickets, ticketCount: tickets.length };
  await pf.files.writeFile("node_format_tickets_output.json", output);
  pf.log.success("Formatted " + tickets.length + " tickets");
  return output;
}

async function node_analyze_tickets(params) {
  const { agentId, message, formattedTickets } = params;
  pf.log.info("Analyzing tickets with AI agent...");
  const result = await pf.mcp.callTool("pinkfish-sidekick", "workflow_agents", {
    action: "invoke",
    agentId,
    message: `${message}\n${formattedTickets}`,
  });
  const parsed = JSON.parse(result.response);
  await pf.files.writeFile("node_analyze_tickets_output.json", parsed);
  pf.log.success("Analysis complete");
  return parsed;
}

global.node_fetch_tickets = node_fetch_tickets;
global.node_format_tickets = node_format_tickets;
global.node_analyze_tickets = node_analyze_tickets;

//---REQUIRED FOOTER - DO NOT MODIFY---
await pf.run(WORKFLOW_NODES, WORKFLOW_EDGES);
//---END REQUIRED FOOTER---

After uploading this code with workflow_update, bind resources:

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_update",
      "arguments": {
        "automationId": "auto_abc123",
        "changeDescription": "Bind Zendesk connection and agent",
        "bindings": {
          "zendesk": "<YOUR_ZENDESK_PCID>",
          "ticketAnalyst": "<YOUR_AGENT_ID>"
        }
      }
    },
    "id": 1
  }'

Setting Up Triggers

Triggers make workflows fire automatically.

Schedule Trigger (Cron)

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_trigger_schedule",
      "arguments": {
        "automationId": "auto_abc123",
        "action": "create",
        "schedule": "0 9 * * *",
        "timezone": "America/New_York"
      }
    },
    "id": 1
  }'

Runs the workflow every day at 9:00 AM Eastern.

API Webhook Trigger

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_trigger_api",
      "arguments": {
        "automationId": "auto_abc123",
        "action": "create"
      }
    },
    "id": 1
  }'

Returns a webhook URL. POST data to that URL to trigger the workflow with input parameters.

Email Trigger

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_trigger_email",
      "arguments": {
        "automationId": "auto_abc123",
        "action": "create"
      }
    },
    "id": 1
  }'

Returns an email address. Emails sent to that address trigger the workflow.

Application Event Trigger

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_trigger_application",
      "arguments": {
        "automationId": "auto_abc123",
        "action": "create",
        "PCID": "<YOUR_SALESFORCE_PCID>",
        "event": "new_lead"
      }
    },
    "id": 1
  }'

Fires the workflow when a specific event occurs in a connected application.

Web Form Trigger

curl -s -X POST "https://mcp.app.pinkfish.ai/pinkfish-sidekick" \
  -H "Authorization: Bearer $PINKFISH_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "workflow_trigger_interface",
      "arguments": {
        "automationId": "auto_abc123",
        "action": "create",
        "schema": {
          "type": "object",
          "properties": {
            "name": { "type": "string", "description": "Customer name" },
            "priority": { "type": "string", "enum": ["low", "medium", "high"] }
          },
          "required": ["name"]
        }
      }
    },
    "id": 1
  }'

Auto-generates a web form. When users fill it out and submit, the workflow runs with their input.

Build Sequence Summary

1. capabilities_discover("search web and summarize with AI")
   -> tools: [search_googlesearch, embedded-groq_generate]
   -> connections: [gmail (pcid_abc), salesforce (pcid_xyz)]

2. capability_details(items: ["web-search", "embedded-groq", "triggers", "resource-bindings"])
   -> Full inputSchema for each tool
   -> Trigger setup instructions
   -> Resource binding patterns

3. workflow_create(name: "Daily News Summary")
   -> automationId: "auto_abc123"

4. workflow_update(automationId, code: "...full workflow JS...", bindings: {...})
   -> Workflow saved with node graph and resource bindings

5. workflow_trigger_schedule(automationId, schedule: "0 9 * * *")
   -> Trigger created — workflow fires daily at 9am

6. workflow_run(automationId)
   -> Test execution — verifies everything works

7. workflow_results(automationId, operation: "read", filename: "output.json")
   -> Inspect actual output data

Triggers API

Platform API

Embedded MCP Servers

Application MCP Servers

Overview

Workflow Tools Reference

How Workflows Work

Node Types

inputSchema Sources

Resource Bindings

The `pf` SDK

Workflow Code Structure

Example: Create and Run a Workflow

Step 1: Discover tools for the task

Step 2: Get full schemas

Step 3: Create the workflow

Step 4: Update with code

Step 5: Run the workflow

Step 6: Inspect the results

Example Script

Complete Code Example: Web Search + Summarize

Complete Code Example: With Resource Bindings and Triggers

Setting Up Triggers

Schedule Trigger (Cron)

API Webhook Trigger

Email Trigger

Application Event Trigger

Web Form Trigger

Build Sequence Summary

Triggers API

Platform API

Embedded MCP Servers

Application MCP Servers

​Overview

​Workflow Tools Reference

​How Workflows Work

​Node Types

​inputSchema Sources

​Resource Bindings

​The pf SDK

​Workflow Code Structure

​Example: Create and Run a Workflow

​Step 1: Discover tools for the task

​Step 2: Get full schemas

​Step 3: Create the workflow

​Step 4: Update with code

​Step 5: Run the workflow

​Step 6: Inspect the results

​Example Script

​Complete Code Example: Web Search + Summarize

​Complete Code Example: With Resource Bindings and Triggers

​Setting Up Triggers

​Schedule Trigger (Cron)

​API Webhook Trigger

​Email Trigger

​Application Event Trigger

​Web Form Trigger

​Build Sequence Summary

Overview

Workflow Tools Reference

How Workflows Work

Node Types

inputSchema Sources

Resource Bindings

The `pf` SDK

Workflow Code Structure

Example: Create and Run a Workflow

Step 1: Discover tools for the task

Step 2: Get full schemas

Step 3: Create the workflow

Step 4: Update with code

Step 5: Run the workflow

Step 6: Inspect the results

Example Script

Complete Code Example: Web Search + Summarize

Complete Code Example: With Resource Bindings and Triggers

Setting Up Triggers

Schedule Trigger (Cron)

API Webhook Trigger

Email Trigger

Application Event Trigger

Web Form Trigger

Build Sequence Summary