workflow engine for AI agents: Teach it once. Every run follows it.

The ConvOps workflow engine turns how the work should go into steps, approvals and detours that every AI agent run follows. Describe it once, in plain words, from the AI client you already use.

no canvas · no pipeline language · one step at a time

example · Claudeconnected to ConvOps

new chat

For every bug: reproduce it, write the fix, then send it to review. If changes are requested, go back to the fix. Wait for my OK, then ship it.
workflow saved · Bug fix · 5 steps · 1 approval · 1 loop
Done. Every new bug now runs it, one step at a time.
example · Bug fixlistening
changes requestedReproduce itWrite the fixReviewsub-workflowYour OKapprovalShip it

example

in short

What is a workflow engine for AI agents?

It stores your process as steps, gates and decisions, and hands your AI agent one step at a time, so the same work comes out the same way every run.

ConvOps is the operations layer for AI agents. It holds your team's process as workflows with gates, a shared memory, scheduled and unattended runs, programs, a plan board, metrics and a full audit trail. It does not sell or host models.

key facts · october 2026

  • In ConvOps, a workflow is a sequence of typed steps with instructions, conditions and gates, stored in your workspace and edited in one place.
  • Your AI client asks ConvOps for the current step, does the work, then asks to advance. The engine checks the step's gate against real state before the cursor moves.
  • A decision step reads the task's fields and picks a path. One path can invoke a whole sub-workflow, nested up to five levels deep; the parent waits, then continues.
  • New tasks are routed to the workflow that owns them by type, vertical, tags, project or a plain-language match.
  • There is no visual canvas and no pipeline language. You describe the process in plain words from the AI client you already use.
  • At Neomanex, a 14 phase plan ran in 7 stages: phases sharing an order ran together as child tasks (October 2026).

best for

  • Processes that drift when they live in a long prompt or instruction file.
  • Teams that want the same steps, checks and approvals on every run, whoever starts it.
  • Agents in Claude Code, Codex, Cursor or ChatGPT that need to follow a process across sessions.

not for

  • Durable execution of application code with retries and timers. Use Temporal for that.
  • Running several steps of one workflow at the same time. Steps run in order; parallel work is child tasks.
  • Designing workflows on a visual canvas.

updated

the problem

A process kept in a prompt drifts.

Long prompts get copied, edited and forgotten. The steps live in someone's head, and the checkpoints are only words.

“What happens when someone leaves?”

The process lived in their head, or in a prompt only they kept up to date.

“Why did it skip the review this time?”

A checkpoint written in a prompt is only wording. A model in a hurry talks its way past it.

“Which copy of the prompt is the real one?”

Everyone keeps their own version. Each one drifts a little, and nobody notices.

try it

Build a workflow. Watch it run.

Pick what your process needs. The workflow re-lays itself, then a run goes through it and stops where you said it should.

example · ConvOps · campaign launchready
new taskBriefDraftReview?Reviewsub-workflowPublish
in plain words · example
  • Brief it, draft it, publish it.
  • Send it to review. If changes are requested, go back to the draft.
compiled
4steps
0approval
1loop
0sub-tasks
one run, start to end
one step at a time

The AI sees only the next step.

The engine holds the cursor. Your AI asks for the current step, does it, and asks to move on. Everything else stays out of its context.

workflow · Bug fixengine holds the cursor
  1. 01Reproducecursor
  2. 02Failing test
  3. 03Fix
  4. 04Review
  5. 05Approve
  6. 06Release
  7. 07Verify
example · Claude · what the AI is holdingstep 1 of 7
in context now
step · reproduce

Reproduce

Recall what the team knows about checkout. Reproduce the crash in Safari and note the exact steps.

not in context
6 other steps stay with the engine
17 words in context · the whole process is 94
example
routing

Work finds its own workflow.

Teach each workflow what it applies to. From then on, new work attaches to the right one when it is created.

new task
type: bugCheckout crashes on Safari
matched by type, tags, project · nobody picks
Bug fix0 attached
applies to · type: bug
Article publishing0 attached
applies to · type: task · tag: article
Credit memo0 attached
applies to · type: task · tag: credit-memo
watch it work

Same work. Same way. Every time.

Example: two bugs, two people, two AI tools, three days apart. Both runs go through the same steps.

01routing

A bug arrives. Its workflow attaches.

Monday. Checkout crashes on Safari. The task is a bug, so the bug-fix workflow attaches on its own. Nobody picks it.

example · ConvOps · Bug fix · runs1 run
Monday · ClaudeCheckout crashes on SafariBug fix attached
Thursday · CodexWebhook returns 500
example · Claudemonday
Checkout crashes on Safari. Can you fix it?
task created · type: bug · Bug fix attached
Claude received · step 1 of 6ReproduceReproduce the crash in Safari and note the exact steps. Then ask to advance.the other five steps are not in the chat
review sub-workflow · returnedChanges requestedThe fix misses the iPad case. The run goes back to Fix, then through review again.loop 1 · back to Fix
in your chat · waiting for approvalShip the Safari fix?
Reply "not yet"Reply "approved"
example · Codexthursday
The payment webhook returns 500. Take it.
task created · type: bug · Bug fix attached
step 1 of 6 · Reproduce
Monday · ClaudeCheckout crashes on Safari6 steps · 1 loop · approved · done
Thursday · CodexWebhook returns 5006 steps · 1 loop · approved · done
02Claude

One step at a time.

Claude asks for the current step and gets only that: reproduce the crash. It does the work, then asks to move on.

example · ConvOps · Bug fix · runs1 run
Monday · ClaudeCheckout crashes on SafariBug fix attached
Thursday · CodexWebhook returns 500
example · Claudemonday
Checkout crashes on Safari. Can you fix it?
task created · type: bug · Bug fix attached
Claude received · step 1 of 6ReproduceReproduce the crash in Safari and note the exact steps. Then ask to advance.the other five steps are not in the chat
review sub-workflow · returnedChanges requestedThe fix misses the iPad case. The run goes back to Fix, then through review again.loop 1 · back to Fix
in your chat · waiting for approvalShip the Safari fix?
Reply "not yet"Reply "approved"
example · Codexthursday
The payment webhook returns 500. Take it.
task created · type: bug · Bug fix attached
step 1 of 6 · Reproduce
Monday · ClaudeCheckout crashes on Safari6 steps · 1 loop · approved · done
Thursday · CodexWebhook returns 5006 steps · 1 loop · approved · done
03review sub-workflow

A detour, then a loop back.

The review detour runs its own steps. Changes requested, so the run returns to the fix, and comes back when the work passes.

example · ConvOps · Bug fix · runs1 run
Monday · ClaudeCheckout crashes on SafariBug fix attached
Thursday · CodexWebhook returns 500
example · Claudemonday
Checkout crashes on Safari. Can you fix it?
task created · type: bug · Bug fix attached
Claude received · step 1 of 6ReproduceReproduce the crash in Safari and note the exact steps. Then ask to advance.the other five steps are not in the chat
review sub-workflow · returnedChanges requestedThe fix misses the iPad case. The run goes back to Fix, then through review again.loop 1 · back to Fix
in your chat · waiting for approvalShip the Safari fix?
Reply "not yet"Reply "approved"
example · Codexthursday
The payment webhook returns 500. Take it.
task created · type: bug · Bug fix attached
step 1 of 6 · Reproduce
Monday · ClaudeCheckout crashes on Safari6 steps · 1 loop · approved · done
Thursday · CodexWebhook returns 5006 steps · 1 loop · approved · done
04you approve

Nothing ships before the approval.

The run stops at the approval gate. Until an explicit approval is given, recorded with your account, the engine does not move the cursor.

example · ConvOps · Bug fix · runs1 run
Monday · ClaudeCheckout crashes on SafariBug fix attached
Thursday · CodexWebhook returns 500
example · Claudemonday
Checkout crashes on Safari. Can you fix it?
task created · type: bug · Bug fix attached
Claude received · step 1 of 6ReproduceReproduce the crash in Safari and note the exact steps. Then ask to advance.the other five steps are not in the chat
review sub-workflow · returnedChanges requestedThe fix misses the iPad case. The run goes back to Fix, then through review again.loop 1 · back to Fix
in your chat · waiting for approvalShip the Safari fix?
Reply "not yet"Reply "approved"
example · Codexthursday
The payment webhook returns 500. Take it.
task created · type: bug · Bug fix attached
step 1 of 6 · Reproduce
Monday · ClaudeCheckout crashes on Safari6 steps · 1 loop · approved · done
Thursday · CodexWebhook returns 5006 steps · 1 loop · approved · done
05a teammate, in Codex

Thursday. Another bug, another AI.

A teammate picks up a different bug in Codex. The same workflow attaches. Same steps, same detour, same approval.

example · ConvOps · Bug fix · runs2 runs
Monday · ClaudeCheckout crashes on SafariBug fix attached
Thursday · CodexWebhook returns 500Bug fix attached
example · Claudemonday
Checkout crashes on Safari. Can you fix it?
task created · type: bug · Bug fix attached
Claude received · step 1 of 6ReproduceReproduce the crash in Safari and note the exact steps. Then ask to advance.the other five steps are not in the chat
review sub-workflow · returnedChanges requestedThe fix misses the iPad case. The run goes back to Fix, then through review again.loop 1 · back to Fix
approval recorded · the run moves onShip the Safari fix?
Reply "not yet"Reply "approved"
example · Codexthursday
The payment webhook returns 500. Take it.
task created · type: bug · Bug fix attached
step 1 of 6 · Reproduce
Monday · ClaudeCheckout crashes on Safari6 steps · 1 loop · approved · done
Thursday · CodexWebhook returns 5006 steps · 1 loop · approved · done
06done

Two runs. One way.

Different people, different AI tools, different days. The work went through the same steps, and every step is recorded on its task.

example · ConvOps · Bug fix · runs2 runs
Monday · ClaudeCheckout crashes on SafariBug fix attached
Thursday · CodexWebhook returns 500Bug fix attached
example · Claudemonday
Checkout crashes on Safari. Can you fix it?
task created · type: bug · Bug fix attached
Claude received · step 1 of 6ReproduceReproduce the crash in Safari and note the exact steps. Then ask to advance.the other five steps are not in the chat
review sub-workflow · returnedChanges requestedThe fix misses the iPad case. The run goes back to Fix, then through review again.loop 1 · back to Fix
approval recorded · the run moves onShip the Safari fix?
Reply "not yet"Reply "approved"
example · Codexthursday
The payment webhook returns 500. Take it.
task created · type: bug · Bug fix attached
step 1 of 6 · Reproduce
Monday · ClaudeCheckout crashes on Safari6 steps · 1 loop · approved · done
Thursday · CodexWebhook returns 5006 steps · 1 loop · approved · done
example · ConvOps · Bug fix · runs1 run
Monday · ClaudeCheckout crashes on SafariBug fix attached
Thursday · CodexWebhook returns 500
example · Claudemonday
Checkout crashes on Safari. Can you fix it?
task created · type: bug · Bug fix attached
Claude received · step 1 of 6ReproduceReproduce the crash in Safari and note the exact steps. Then ask to advance.the other five steps are not in the chat
review sub-workflow · returnedChanges requestedThe fix misses the iPad case. The run goes back to Fix, then through review again.loop 1 · back to Fix
in your chat · waiting for approvalShip the Safari fix?
Reply "not yet"Reply "approved"
example · Codexthursday
The payment webhook returns 500. Take it.
task created · type: bug · Bug fix attached
step 1 of 6 · Reproduce
Monday · ClaudeCheckout crashes on Safari6 steps · 1 loop · approved · done
Thursday · CodexWebhook returns 5006 steps · 1 loop · approved · done
checkpoints

Put a checkpoint where it earns its place.

A checkpoint can wait for an explicit approval, a metric or a report. Place them where you want them, or run a workflow with none.

reproduceClaude
fixClaude
reviewdetour
approveapproval
releaseClaude
verifydone
the graph

See the whole workflow at once.

The graph view draws every step, decision and approval from start to end. Each sub-workflow opens in place, right where the work takes the detour.

example · workflow · graphbug fix
Reproduce
Failing test
Fix
Reviewinvokes Code review
Approveapproval
Release
Verify
sub-workflow · Code review · called by Review
Code review
Read the diff
Run checks
Changes?yes: back to Fix
Backto Approve
  • Every path, labelled.Each decision shows where each answer goes, and a loop back to an earlier step is drawn as its own path.
  • Sub-workflows open in place.A detour opens beside the step that calls it, with its own steps and decisions. Close it again with one click.
  • Click a step to read it.The step's instructions and its agent open in a side panel.
the shapes

Five shapes cover most processes.

Steps

Each step carries its own instructions, handed over only when the work reaches it.

one at a time

Gates

An approval, a required metric or report, finished child tasks: checked on every move. Not met, the run does not go forward.

it refuses

Detours

A decision can send the work through a whole sub-workflow. It always comes back.

loops included

Parallel

Child tasks that share an order start together. The parent waits for all of them.

start together

Routing

New work is matched to the workflow that owns it, by type, tags or project.

nobody wires it

versions

Change the process. Keep running work steady.Planned

Every publish of a workflow is a version. Running work keeps the version it started with, and you can roll back at any time.

Versions

Each publish is a numbered version of the whole workflow, the sub-workflows it calls included.

whole workflow

Drafts

Edit a draft while running work keeps the published version. Publish when it is ready.

publish when ready

Pinned runs

A run stays on the version it started with. Move running work to a new version when you choose, with a preview first.

you choose when

Rollback

Set an earlier version live again. Nothing is deleted.

history kept

Today every run follows the live workflow. Agents are already versioned: each step runs its agent's active revision, and teams pin the revisions they tested. How agents are versioned

the difference

A long prompt vs a compiled process.

example · bugfix-prompt-v7-final (2).md4 copies

IMPORTANT: always ask before releasing.

IMPORTANT: always ask before releasing.

a sentence, not a gatethe whole process, every message

  • The entire process arrives at once, up front
  • Instructions go stale silently, per copy
  • The model juggles every constraint at all times
  • Checkpoints are wording a model can talk past
  • The process lives in whoever wrote the prompt
example · ConvOps · Bug fixone step in context
reproduceClaude
fixClaude
reviewdetour
approveapproval
releaseClaude
verifydone

checked by the engineexample

  • One step arrives at the moment the work reaches it
  • Edit the step once; every future run inherits it
  • The model sees only what is next
  • Gates are evaluated by the engine and refuse
  • The process lives in the template, with an audit trail

Routing

A template declares what it applies to: task type, vertical, tags or project. Matching work attaches the workflow at creation.

payload
{
  "applies_to": [{
    "type": "bug",
    "vertical": "development",
    "default": true
  }]
}
Three tasks arrive. Only one belongs here.

Steps

Your client calls workflow_advance over MCP and receives the current step only. Edit a step once and future runs inherit it.

payload
{
  "id": "failing-test",
  "name": "Failing Test",
  "status": "in_progress",
  "instructions": "Write the test that
     reproduces the crash. Red first.
     Notify me when it goes green."
}
Step 2 of 7. The other six stay out of context.

Gates

Evaluated server-side on every advance: requires_approval, wait_for_children, sequential_children, task_status, require_context, plus any required metric or report bound to the step. A miss returns unmet_conditions and the cursor stays.

payload
{
  "current_step": "approve",
  "unmet_conditions": [
    "Requires explicit user approval"
  ]
}
A refused gate, verbatim.

Decisions and loops

A decision records which condition matched. A branch is an invoke into a sub-workflow; on the way back, resume_at can point backward. That is the fix loop. There is no forward goto.

payload
{
  "decision": true,
  "conditions": [
    { "id": "simple",
      "when": { "task_field": "complexity",
                "equals": "S" } },
    { "id": "unclear",
      "invoke": "deep-research" },
    { "id": "go", "when": "default" }
  ]
}
Same task, same rules, every single run.

Parallel work

Child tasks that share an order value start together; the rest start one after another. Each child can target its own executor, and the parent's wait_for_children gate holds until all are done. A step can also tell your AI session to start several agents at once.

payload
{
  "children": [
    { "title": "fix", "order": 1 },
    { "title": "tests", "order": 1,
      "executor": "your-runner" },
    { "title": "docs", "order": 1 }
  ],
  "parent_gate": { "wait_for_children": true }
}
The intelligence is replaceable. The process is not.

Authoring

Templates live in the database, built from steps, reusable fragments and workspace-wide globals. Edits are surgical, dry-run validated, and runs in flight survive them.

Close the loop

Agents record metrics from their runs. A workflow reads them and rewrites workflows, agent instructions or model tiers, and a gate can require the metric before the next step. People set the metrics and the targets. Add an approval gate to hold a change until an explicit approval is given, recorded with the account that gave it. Example: a support team reads first-reply quality each week and tightens the triage step.

questions

Workflow engine questions, answered.

What is a workflow engine for AI agents?

A workflow engine for AI agents holds a process as structure and hands it to the agent one step at a time. ConvOps stores each workflow as steps, gates and decisions in your workspace. Claude Code, Codex, Cursor or any MCP client asks for the current step, does it and asks to advance. Operations, finance, support and engineering teams use it so the same work runs the same way every time.

Is this a prompt library?

No. ConvOps is not a prompt: a prompt hands the model the whole process at once and hopes it is followed. In ConvOps the engine holds the cursor: steps arrive one at a time as the work reaches them, gates are checked server-side on every advance, and a gate that is not met returns its unmet conditions instead of moving.

Can a decision branch the flow?

Yes. A ConvOps decision step checks its conditions against real task fields and records which one matched. A branch is a detour: the condition invokes a sub-workflow, which runs to the end with its own steps and gates and then returns. The main line stays linear by design, so every run reads as one auditable line.

How do loops work?

On the way back from an invoked sub-workflow, ConvOps can resume at an earlier step named by resume_at. That is the fix loop: a review finds problems, the run goes back to the fix step, and the cycle repeats until the work passes. A credit memo sent back by legal review follows the same loop as code sent back by a reviewer.

Can steps run in parallel?

Yes, two ways. A step's instructions can tell your AI session to start several agents at once: ConvOps holds the step, your session does the fan-out. Or split the work into child tasks: children that share an order start together, each can target its own executor, and the parent's wait_for_children gate holds until all are done.

Where do workflows live?

In the ConvOps database, scoped to your workspace: steps, reusable fragments and workspace-wide globals. You install starting points from the template catalog and edit them in conversation over MCP from your own client. A review sequence written once as a fragment can be reused by every workflow that needs it.

What happens to runs already in flight when I edit a template?

They keep going. ConvOps applies template edits surgically and validates each one with a full dry run before it lands. Runs already in flight continue against the updated definition, so a fix to a step reaches the next run that gets there without restarting anything.

Can a run stay on the workflow version it started with?

No. ConvOps does not version workflows today: every run follows the live workflow. Agents are versioned: each save of an agent's instructions is a revision, teams pin the revisions they tested, and an earlier revision can be made active again.

How does a new task get its workflow?

Routing. ConvOps templates declare what they apply to, by task type, vertical, tags or project, and matching work attaches its workflow at creation. A shared-services team can send IT, legal and access requests to their own workflows this way, and nobody wires anything per task.

Can a workflow improve from its own numbers?

Yes, when you build it that way. In ConvOps agents record metrics from their runs, and a workflow can read them and edit workflow steps, agent instructions or model tiers. A gate can require a metric before the next step moves. People set the metrics and the targets. Add an approval gate to hold a change until an explicit approval is given, recorded with the account that gave it.