Should you choose Temporal or ConvOps?
Choose Temporal when engineers need code that survives failures. Choose ConvOps when the agents you already use must follow a process with sign-off, memory and a record.
choose Temporal if
- Engineers are building a service that must survive crashes, retries and long waits.
- The process is code: payments, provisioning, data pipelines, an agent loop you own.
- You need guarantees at scale and a choice of eight language SDKs.
- You are happy to run workers, or pay Temporal Cloud by usage.
choose ConvOps if
- The "workers" are AI clients and people, not services you deploy.
- Process owners should change the steps without a code release.
- You need approval gates, shared memory and an audit trail for agent work.
- You want to start attended in your AI client and go unattended later.
How do Temporal and ConvOps compare?
Side by side, as of October 2026. ConvOps is the operations layer for AI agents: an MCP server that holds the process your AI clients follow, with approval gates, a shared memory and an audit trail. Every Temporal cell links to the official source it comes from; ConvOps cells describe the product as it ships today.
| Capability | Temporal | ConvOps |
|---|---|---|
| What runs the steps | Your workers run workflow and activity code. Activities touch the outside world: APIs, databases, LLM calls. [1] | Your AI client runs each step. ConvOps hands it one step at a time and runs no AI models itself. |
| Where the process lives | Workflow code in your repository, in .NET, Go, Java, PHP, Python, Ruby, Rust or TypeScript. [1] | A workflow in your ConvOps workspace, reached over MCP and changed by asking your AI client. |
| Human approval gates | A running workflow receives Signals and Updates. Temporal's AI cookbook shows an agent that waits for human approval by Signal. [2] | Approval gates on any step, evaluated by the engine. The run waits at the gate, for minutes or days, until approval is passed. |
| Memory | Not an agent memory feature: workflow state lives in your code's variables, made durable by the Event History. [1] | A shared brain of memories and routes, recalled across sessions, AI clients and workspaces. |
| Audit trail | The Event History: a complete, ordered record of everything that happened in a workflow execution. [1] | An audit row for every change to tasks and workflow runs, keeping the agent's self-declared label apart from the server-verified human. |
| Attended or unattended | Unattended by design: once a workflow starts, it runs to completion, whether that takes a second or a year. [1] | Attended in your AI session, or unattended through executors and RRULE schedules, including one pod per run on your own Kubernetes. |
| Self-hosting | Run the open-source, MIT-licensed Temporal Service yourself, or use Temporal Cloud. [5] | Hosted at mcp.convops.app, or self-hosted on your Kubernetes with one Helm chart on Enterprise. |
| Pricing model | Self-hosting is free. Temporal Cloud is pay as you go from $50 per million actions plus storage, with $150 in free credits for 90 days. [4] | Free $0, Pro $79/mo ($66/mo billed annually), Business $299/mo ($249/mo annually), Enterprise from $2,500/mo. Every feature on every plan; you pay for scale. Your AI client is billed separately. |
| Client support | Not an AI client integration: your code calls it through the SDKs. [1] | Any MCP client. Verified: Claude Code, Claude (web, desktop and Cowork), OpenAI Codex, ChatGPT, Cursor and OpenCode. |
What is Temporal?
Temporal is a durable execution platform. A workflow is your business process written as ordinary code, and once it starts, it runs to completion through crashes, deploys and long waits.
Workflows hold the control flow; activities do the work that touches the outside world, such as calling an API or writing to a database. Every step is recorded in the Event History, which is what lets a workflow resume exactly where it was. Running workflows can receive Signals and Updates and answer Queries while they run, and there are SDKs for eight languages (understanding Temporal).
For AI, Temporal publishes guides and cookbook recipes, including a human-in-the-loop agent that proposes an action, waits for approval through a Signal, and then executes or cancels (AI cookbook, Temporal for AI).
Is ConvOps a Temporal alternative for AI agents?
They solve different problems. Temporal makes your code durable. ConvOps makes an AI agent's process governed. The overlap is the word "workflow".
A Temporal workflow is code an engineer writes and deploys; the people in it act through Signals your application sends. A ConvOps workflow is a process your AI client walks one step at a time, written by describing it, and the people in it decide at approval gates in the same conversation.
If you are building an agent service, Temporal is the right substrate and ConvOps is not trying to replace it. If you want Claude Code, Codex or ChatGPT to follow your team's process with sign-off and a record, writing Temporal code for that is a heavy way to get there.
Where is Temporal stronger?
In execution guarantees for code. ConvOps is not a code runtime.
Durable execution
Workflows survive process crashes and resume from the Event History, for a second or a year.
Polyglot SDKs
Eight languages, with workflows as ordinary loops, conditionals and function calls.
Activities with the outside world
API calls, database writes and payments with Temporal handling the failure modes.
Scale and operations
Self-host the open-source service, or hand its operation to Temporal Cloud.
Where is ConvOps different?
It is for processes where the worker is an AI client or a person, and the owner is not necessarily an engineer.
No code to write or deploy
Describe the workflow to your AI client. Change one step by asking.
Gates people act on directly
An approval gate holds the run until approval is passed on the advance, from the AI client you are already in.
Persistence for agent work
Every step is persisted, so any session or AI client resumes the run at its current step.
Memory and audit built in
A shared brain across clients, and an audit row per change with the verified human.
ConvOps does not retry your functions, run timers inside your code or guarantee exactly-once side effects. If your process needs those, it needs Temporal or something like it.
Can you use Temporal and ConvOps together?
Yes, at different layers. Temporal runs the durable service; ConvOps governs the agent-driven process around it, such as building and releasing that service.
A team building on Temporal can run its engineering process in ConvOps: change request, design approval, implementation in Claude Code, tests against a local Temporal dev server, release sign-off. We do not ship a Temporal integration; the two meet in the work your agents do.
one ConvOps workflow, with Temporal inside one step
How do you start with ConvOps for agent workflows?
Connect your AI client and describe the process. There is no worker to deploy.
- 1
Connect your AI client
Claude Code takes one line; Cursor, Codex, ChatGPT and others take the server URL in their MCP settings.
claude mcp add --transport http convops https://mcp.convops.app/ - 2
Describe the process
Name the steps, the approval gates and what happens on a failed review. Your AI client creates the workflow.
- 3
Run it attended first
Walk a few tasks through it in your own session. Approvals happen where you work.
- 4
Go unattended when it has earned it
Attach an RRULE schedule and an executor. Gates still hold the run for a person. See run Claude Code unattended on Kubernetes.
What else do people ask about Temporal and ConvOps?
Short answers to the questions that come up most, with the same sources as the table above.
Is Temporal good for AI agents?
For agent services you build in code, yes. Temporal gives durable execution, Signals for human input and an Event History, and it publishes AI guides including a human-in-the-loop agent recipe. It is a code platform, so engineers write and operate the workflows.
Does ConvOps provide durable execution?
Not for code. ConvOps persists every step of a workflow run, so any session can resume it at its current step. It does not execute or retry your functions.
Do I need engineers to use ConvOps?
No. Workflows are created by describing them to an AI client. Engineers are useful for executors and self-hosting, not for writing the process.
Can both be self-hosted?
Yes. The Temporal Service is open source and can be run yourself, or you can use Temporal Cloud. ConvOps Enterprise runs on your Kubernetes with one Helm chart.
How is pricing different?
Temporal Cloud charges by usage, from $50 per million actions plus storage, and self-hosting is free. ConvOps has a free plan, Pro at $79 a month ($66 billed annually), Business at $299 a month ($249 annually) and Enterprise from $2,500 a month, and you bring your own AI client.
Where do these facts come from?
From the official documentation, repository and pricing pages for Temporal, checked in October 2026. Anything we could not verify there is left out.
- [1]Temporal docs: Understanding Temporal
- [2]Temporal AI cookbook: Human-in-the-loop AI agent
- [3]Temporal docs: Durable AI
- [4]Temporal Cloud pricing
- [5]Temporal Service repository and MIT license on GitHub
Spotted something out of date? Tell us and we will correct the page and its date.