Workflows
Chain AI prompts, web research, and your connected apps on one canvas. Publish a deployment and run it on a schedule, from a webhook, in your own code, or straight from the AI assistant you already work in.
Free Developer plan · No card required · Templates you can clone in one click
Start
AI Prompt
Web Research
Condition
Slack
End
01
ANATOMY
Start, steps, End. Every run follows the graph you drew.
A workflow is a fixed path from Start to End. Each step does one job and hands its output to the next through workflow variables, so the run does the same thing every time it is triggered.
Start inputs
01
Define the variables a run needs, such as a topic, a URL, or a list. Every trigger fills the same inputs, so the workflow behaves the same wherever it starts from.
Steps
02
Add a step with the + connector between any two nodes. Each step does one job and passes its output forward as a variable the next step can read.
Branches and loops
03
A Condition step splits the path on a rule. Iteration loops run the same steps for every item in a list, then merge back before End.
End and output
04
End collects what the run produced: text, structured data, a generated file, or a message already delivered to Slack, Notion, or a sheet.
02
BUILD
Drag, connect, and run a single step.
The editor is a canvas, not a form. Insert, reorder, duplicate, or disable steps while you work, and open any step’s settings beside the graph instead of on top of it.
Every change stays on the canvas.
Click the + connector on any edge to insert a step, or drag a step card onto a new insert point to move it. Branches, merge points, and loop ends all accept drops, so restructuring a flow never means rebuilding it.
Watch the run while it is still running.
Run Workflow opens Live Progress, you fill in the Start inputs, and each step reports as it completes. Expand a finished step to read its output before the rest of the run is done, then change a setting and run again.
03
DEPLOY
Publish once. Start it from wherever the work begins.
A deployment is the runtime target for a workflow. Schedules, webhooks, and your code all call a published deployment variant, never the draft you are still editing.
Schedules
Run a deployment variant daily, weekly, or on your own recurrence, with a timezone and saved Start inputs. Structured settings only, no cron syntax.
Webhooks
Give Zapier, your CRM, or any other service a URL. The deployment secret authenticates the call, and the result comes back as a signed callback.
API and SDKs
Invoke a deployment with your API key from cURL or the Python, Node.js, Ruby, and PHP SDKs. Wait for the response or get it delivered to a callback URL.
Run by hand
Start a run from the editor or the deployment page and type the inputs in. Useful for one-offs and for checking a change before a schedule picks it up.
webhook trigger · weekly-research-brief
# start the published variant from any system that can POST
$ curl https://api.fetchhive.com/v1/workflow/webhooks/<endpoint>/variants/<variant> \
-H ‘X-Fetch-Hive-Webhook-Secret: <deployment secret>‘ \
-d ‘{ “inputs”: { “topic”: “State of enterprise AI” },
“async”: { “callback_url”: “https://example.com/callback” } }’
✓ queued run_01HZ… for weekly-research-brief / default
Status
running
Inputs
topic = “State of enterprise AI”
Callback
signed with X-Fetch-Hive-Signature (HMAC SHA256)
Result
delivered to https://example.com/callback
VERSIONS AND DEPLOYMENTS
A version is an editor checkpoint. A deployment variant is what the outside world calls. Keep editing and testing the draft while the published variant keeps running exactly as it was when you shipped it.
Save a named version whenever you reach a checkpoint
Compare history and roll back without touching the live deployment
Variants run the same workflow with different settings or inputs
Schedules and webhooks run the published variant, never the draft
04
INTEGRATE
Connect your apps once under Integrations and any workflow step can use them. 73 live providers, more than 6,400 agent tools, and every call recorded in the run so the output lands where your team already works.
6,400+
AGENT TOOLS
Actions your steps and agents can call today
73
LIVE PROVIDERS
Apps connected once, reused by every workflow
40+
STEP TYPES
AI, research, utilities, logic, and app actions
Source documents
Input
Research workflow
Runs in Fetch Hive
Brief in Google Docs
Output
Slack notification
Delivered
05
CONNECT
Run workflows from the AI assistant you already use.
Fetch Hive ships an MCP server. Connect Claude, ChatGPT, Cursor, or Codex once, and your assistant can find, build, schedule, run, and approve workflows in plain language. Every run is logged like any other.
your assistant · fetch hive mcp
connected
Run the weekly research brief on “State of enterprise AI” and post it to #marketing.
search_workflows
get_workflow
Found weekly-research-brief, default variant. It takes one input, topic, and ends with a Slack step. This run is quoted at 12 credits. Start it?
Yes. And run it every Monday at 9am from now on.
run_workflow · approved
get_run
schedule
Run run_01HZ… finished in 48 seconds and the summary is in #marketing. A Monday 09:00 schedule is now saved on the deployment with the same input.
run_01HZ… · completed · 48s · 12 tasks · logged in Workflow Logs
What your assistant can do once it is connected
Find and read any workflow in your workspace
Start a run with the cost quoted first, then approve it in the chat
Clone a template into your workspace and change the steps
Deploy, schedule, and share a workflow as a public link
Approve or deny a paused Human in the Loop step
Open the run trace and final output without leaving the chat
WORKS WITH
Sign in once with OAuth. Nothing runs without your yes.
06
OBSERVE
Every run leaves a trace you can open.
Logs keep each run with its status, the step-by-step trace, the final output, and the webhook activity around it. A failed run is one click away from the step that failed.
01
The task
Produce a company research brief with sources. The input is a company URL; the expected output is a one-page document your team can read.
02
The execution
Every model call, tool call, duration, and recorded cost. Open a meaningful step to see its input and output.
03
The output
A readable excerpt from the resulting brief, with sources where supported, next to the run summary.
Every run records:
The input and output of every step, including loop passes
Model calls, tool calls, duration, and recorded cost per step
Approval decisions, who made them, and when
Failures, with the step that caused them one click away
Value is a human judgement. The run gives you the step trace and the recorded cost to make it, and the same trace is what your assistant reads over MCP.
07
START FASTER
Clone a template, or share a workflow of your own.
Most workflows start as a copy of one that already works. When yours is ready, a public link lets anyone run it without an account.
Templates · Free to clone
I WANT TO
Browse workflows built by Fetch Hive and the community. Clone one into your workspace, change any step, and it is yours.
Filter the marketplace to workflows only
Keeps the full step graph, branches and loops included
Every template shows how many people use it
No account needed to browse
Public links
I WANT TO
Turn a deployment into a public link with its own title and description. People run it without an account, and the runs are billed to your workspace so you stay in control.
Edit the copy without redeploying the workflow
Test the share yourself before you send the link
Switch it off at any time
Free Developer plan
MOST USED WORKFLOW TEMPLATES
PART OF THE FETCH HIVE PLATFORM
Connect Slack, Notion, Google, HubSpot, and the rest once under Integrations, and every workflow step can use them. The same account runs the MCP Gateway and the templates marketplace.
FAQ
Everything else is in the workflow docs.
A workflow runs a fixed graph of steps you drew, in the order you drew it. Use one when you need several model calls or tool actions in a known sequence with predictable output. Open-ended goals are a job for agents, which have their own page.
