Official account for FetchRelay. Create AI workflows for your customers in a jiffy. Give each customer their own rules and run workflows from your app.
Pick one case your team handled by hand today.
Put a sample into FetchRelay, run a preview, and inspect the JSON your app would get back.
New workspaces get 10 free runs. https://t.co/nj1QSmVRzs
Want to see what goes into a workflow before signing in?
Open a FetchRelay starter and inspect the job, instructions, and expected result.
Sign in when you’re ready to save it and run a preview. https://t.co/nj1QSmVjJU
An urgent job has zero qualified technicians available.
A run can finish with output.outcome set to needs_human. Your app can use that field to route the case.
See the result format. https://t.co/W1yeblRAhF
What would happen if you approved a schedule change during a test?
FetchRelay simulates approved writes in preview. You can inspect the proposed change and see the result the run returns.
Test approval and rejection before you publish.
A shipment misses its promised delivery window.
Have the workflow check the latest tracking update and return the revised ETA, its source, and whether someone needs to step in.
Try a sample where the carrier hasn’t provided a new ETA.
A customer confirms that technicians should use the loading dock after 6 p.m.
Enable customer memory for details like that so future runs for that customer can use them.
Preview the workflow with a sample evening appointment.
A dispatcher needs a recommendation and will make the schedule change themselves.
Give the workflow tools to read the job and schedule. The harness only exposes the tools enabled for that deployment.
Run a sample and inspect which tools it called.
The model sends a technician ID to a tool that expects a job ID.
FetchRelay checks the tool arguments against its schema before executing the call.
Use distinct fields such as job_id and technician_id when you define a tool.
One customer needs to pause their rollout for the weekend.
Pause their deployment in Studio. Other customers can keep running the same workflow version.
Resume that deployment when the customer is ready.
You’re changing how a workflow escalates urgent jobs.
Edit the draft while customers keep using the published version. Publish the update when it’s ready.
Assign it to one customer deployment first and inspect the next run.
Your scheduling tools are on a private MCP server.
Run FetchRelay’s outbound connector inside your network. It picks up tool requests and sends the results back.
See the connection setup. https://t.co/8TcG69lC2G
A dispatcher approves a change an hour after the workflow proposed it.
FetchRelay saves the pending action and run state while it waits. The workflow resumes when the decision arrives.
See how to send the decision from your app. https://t.co/LEnxLesq2Z
Your backend times out after starting a job.
Retry with the same Idempotency-Key, deployment ID, and input. FetchRelay returns the original run.
Use one key per business event. https://t.co/RTz7eFNyZg
A recommendation surprised your dispatcher. Where did it come from?
Open the run to see its input, which tools it called, and what it returned.
Start with the record that informed the decision. https://t.co/W1yeblRAhF
An invoice says 12 units. The receiving record says 10.
Ask the workflow to return the mismatched line, both source records, and who should review it.
Try the invoice starter. https://t.co/ceeiKXy9me
A dispatch result needs to fit into your app’s existing job view.
Require fields such as job_id, dispatch_status, and next_action. FetchRelay checks completed results for them.
Open the structured result after a preview to see what your backend receives.
Want to test dispatch when the schedule is full?
Put a sample schedule in _toolFixtures under the schedule tool’s name. FetchRelay gives the model that response in preview.
Check when the workflow hands the job back to a dispatcher.
Two customers use the same scheduling workflow. Their buildings have different access rules.
Add each customer’s access instructions to their deployment.
Send the same sample job to both and compare the recommendations.
The model proposes moving a service visit to 8 a.m.
With write approval enabled, FetchRelay pauses and shows the proposed change for review.
Your app can send the decision back to the run. https://t.co/LEnxLerSdr
A dispatch job may need to check the schedule, skills, and parts availability.
POST /v1/runs queues the job and returns a run ID so your backend can check progress.
Copy the request. https://t.co/RTz7eFNyZg