@aToy0m0 Please let us know if you run into any other pitfalls and we'd love to make Claude Managed Agents even easier to use. Thanks again for writing the article! :)
Been noodling on how I feel about AI writing and music. I think this is the essence of where I'm landing:
Selection is authorship.
Curation is a fingerprint.
Judgement and taste are more important than ever.
@levelsio Thanks again for the feedback, Pieter!
We just shipped an update so you can click on Auto-reload in that home view and manage your auto reload settings.
@levelsio Also lmk if you ever want to chat about prompt caching! I am not 100% sure but your products might all be mostly unique one off prompts, but if any are multi turn, prompt caching can likely save you some money.
@levelsio Hey Pieter 👋
We can work to make that easier to find.
If you go to https://t.co/aly8LzACUy
Then click on the "Auto reload on" link and go through the steps you can disable.
@testingcatalog ant is a little different from the Claude Code CLI. Claude Code is a full harness, the ant CLI is a command line interface to use the API. For example using the messages API or automating spinning up a team of Claude Managed Agents.
@ojabowalola Many Stainless users generated CLIs similar to ant. ant was originally generated with the Stainless generator and has several additional helpers on top of the REST API wrappers.
The ant CLI is a super clean wrapper around the Claude Platform APIs. It also has a ton of helpers for working with Managed Agents. For instance, you can use `ant beta:worker` to poll and run jobs when self hosting for agents.
We’ve added a CLI for Claude Platform to make every API endpoint runnable from your terminal.
Call the Messages API, stand up Claude Managed Agents, pipe results straight into your shell.
The ant CLI is well understood by coding agents (Claude Code) using the claude-api skill.
@JohnGregQuantum@ClaudeDevs You could totally have these buttons fire off events to create new sessions, approve, interrupt etc.
Could probably even have it use `ant beta:worker` poll and run to have the device run the self-hosted environment for the agent!
Also definitely finding myself pairing yaml files that define the agent, environment, vaults, memory etc. and then using the CLI to quickly spin up Claude Managed Agents that I can enlist to take longer running work off my plate.
We’ve added a CLI for Claude Platform to make every API endpoint runnable from your terminal.
Call the Messages API, stand up Claude Managed Agents, pipe results straight into your shell.
The ant CLI is well understood by coding agents (Claude Code) using the claude-api skill.
I think the ownership question is right but the list is off Managed Agents has tool-call approvals, credential vaults, and a fetchable event log. Also built in evals with the new Outcomes feature
What it can't give you today: ZDR/HIPAA coverage (sessions are stateful by design), and it's still beta.
With the SDK, model generated code runs on infra you have to sandbox and operate.
@zantrati@ClaudeDevs Can you explain what you're running into? If you're able to share more detail we might be able to point you in the right direction.