Niches · Workflow automation

Find out why a run failed and recover it without repeating its effects

No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.

0companies, full fit
13companies, fit with a limit
47companies carry the tag
13share of the category's work
8first-hand accounts

What counts as done

The agent names the failing step and whether the failure is transient or a broken connection from logged attempt data, and a recovery run completes with zero duplicated external writes.

Measured, for our own tests: this takes under 5 calls. No docs we read state these figures, so they don't count against a listing here.

How agents ask for it: why did last night's run fail, retry the failed execution, did the emails go out before it broke, is my workflow still running

Does the job, with a limit 14

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

ListingWhy
Activepieces API + MCP C
Activepieces Inc.
ap_list_runs and ap_get_run give per-step outputs, errors and durations, ap_list_connections shows ERROR status, and ap_retry_run FROM_FAILED_STEP keeps earlier outputs. No idempotency key, and docs don't say what a failed loop step repeats. Run data is kept 30 days. source
Microsoft Copilot Studio MCP D
Microsoft Corporation
list_runs, get_run, list_run_actions, get_run_action_content and inspect_run read per-action detail, but there is no resubmit or retry action, and the README says never to retry an uncertain write automatically. source
Gumloop C
AgentHub Inc. (doing business as Gumloop)
Retrieve session returns messages and state, and get_run_history is an MCP tool, but no retry or resume endpoint is documented, so recovery is a new session that may repeat writes. source
Kestra C
Kestra Technologies SAS
GET /executions/{id} plus actions/restart (failed tasks, chosen revision) and actions/replay from a task run id cover diagnosis and recovery, but no idempotency key, so a partly written failed task repeats. source
Kognitos C
Kognitos, Inc.
Runs, run events and exceptions with troubleshooting guides name the failure, and :replay reuses recordings of completed book calls rather than repeating them. Replay works only for single-phase (Main) automations and takes updated code. source
Make API + MCP B
Celonis, Inc.
Execution details, module logs and incomplete executions (bundles, failed blueprint, resolve attempts) name the failing module. Replay reruns the latest scenario with the original data, repeating earlier writes, and no idempotency key is documented. source
n8n API + MCP D
n8n GmbH
Executions keep per-node data (acc-01) and the retry POST takes an Idempotency-Key, but both execution retry and Retry On Fail resent 49 delivered emails, the latter showing success (acc-31). source
Paragon ActionKit + MCP D
Forge Technology, Inc. (d/b/a Paragon)
The Task History API lists executions by user, workflow and FAILED status, but per-step input and output and a replay call aren't in it, and auto-retry repeats a whole step. source
Pipedream API + MCP BB
Pipedream, LLC
Event history exists for Workflows until shutdown, but action runs have no idempotency and acc-30 records a timed-out SSH write that left a file with the wrong checksum. source
Microsoft Power Automate C
Microsoft Corporation
FlowRun gives each run's status, trigger type, error code and error message, but at run level, not per action, and no documented call resubmits a run. source
Prismatic C
Prismatic Software Inc.
replayExecution and bulkReplayExecutions re-run an execution, but they take no idempotency key and the docs read don't say whether completed steps repeat. source
Tray.ai API + MCP C
Tray.ai, Inc.
Headless lists a workflow's executions, reads one and drills into each step's input and output, but no replay or idempotency key appears in the pages read, so a re-fire repeats every step. source
Windmill API + MCP D
Windmill Labs, Inc.
Runs show status, logs, inputs and steps, restart/f/{id} restarts a flow at a chosen step keeping earlier iterations, and a caller-supplied job_id refuses a duplicate. Whether per-step retry attempts are readable wasn't checked. source
Workato API + MCP C
Workato, Inc.
Get a job returns lines with each step's input, output and error. Repeat jobs reruns whole master jobs on the latest recipe version, 25 a call, with no idempotency key, so steps that succeeded repeat. source

Doesn't do this job 4

Listings in the category that do something else, with the reason.

ListingWhy
Bind AI F
1Fort Inc.
No run or execution history endpoint in the API definition. source
Cooper (Mandl) F
Cooper AI Tech, Inc.
No run history is exposed outside the workspace. source
Motive Agents F
Motive Technologies, Inc.
No run history for Agents is exposed through the developer portal. source
Node-RED B
OpenJS Foundation
No execution history, per-run data or retry in the Admin API; failures show only in logs and debug output. source

What stops agents finishing

  • retry re-sends items that already succeeded and reports success
  • no idempotency key on run or retry
  • error analysis tool invents an error on a successful run
  • execution not saved for failed production runs by default

Around it

Before: workflow-automation.unattended-sync, workflow-automation.run-workflow-as-tool. After: workflow-automation.connect-app-accounts, chat.report-to-user.

Where this comes from

Weight 13 of 100 in Workflow automation: 8 of 48 accounts, all n8n (forum threads on retries and liveness, n8n-mcp executions tool, Hacker News); other listings unevidenced

The same as data: /api/v1/use-cases/workflow-automation.debug-failed-run.json, or the MCP tool get_use_case with id workflow-automation.debug-failed-run. Other jobs in this category: /api/v1/niches?category=workflow-automation.

Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person. A mark says what the docs show, not what a run measured.

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.