Developer Experience at @AnthropicAI
Joined April 2009
- Tweets 6,117
- Following 934
- Followers 3,033
- Likes 10,895
789 Photos and videos
Going to do my best to share helpful updates about how to get the most from the Claude Platform: APIs, SDKs, the ant CLI and more.
If you have any feedback about existing API features you’d love to see, I’m all ears!
Pumped to be back in a position where there’s so much I’m excited to use and hype.
9
480
CJ Avilla retweeted
Jun 13
We've reset 5-hour and weekly rate limits for all users.
1,732
1,073
22,157
2,469,781
One little but great quality of life thing I've wanted for a long time just landed in @1password dev environments: sorting by key name 😅
I learned a better way to handle .env files with @1Password environments and wanted to share!
youtube.com/watch?v=Q3dAjue6…
1
272
I wanted to share some Claude Managed Agent tutorials for self-hosted sandboxen to help you get started faster.
It’s pretty sweet how you can have managed agents run code and use tools wherever you want to call “home” as your self-hosted sandbox. (I have so many demo ideas!!)
Here’s a growing list of tutorials that I’ll pile onto as more are built:
Jun 12
Claude Managed Agents can operate in a sandbox you control, on your own infrastructure or with any provider you choose.
Today we added new guides for @blaxelAI, @e2b, @googlecloud, @namespacelabs, and @superserve_ai, so you can choose the best fit for your use case.
2
4
42
3,950
.@daytonaio cookbook ---
A FastAPI webhook drains the queue, creates a Daytona sandbox per work item, uploads sandbox_runner.py and starts it. Full Linux container so the default toolset works as-is. Host it anywhere that can reach the Daytona API; pre-bake the SDK to skip the ~10–15s pip cold start.
github.com/anthropics/claude…
1
2
77
.@vercel_dev
A Vercel Function verifies drains, then spawns a Vercel Sandbox running the TS EnvironmentWorker.handleItem() against a real filesystem.
Optional Vercel KV maps session_id → sandbox_id for reuse across turns; without it every turn is a fresh sandbox (dedup makes that wasteful, not wrong). /workspace is ephemeral — no volume API.
github.com/anthropics/claude…
2
70
📢 you have to pass a new arg to tell the API which model to fallback to!
Don’t just update your model ID to fable, also pass
fallbacks: [{ model: ‘claude-opus-4-8’ }]
Jun 9
Claude Fable 5 is our first generally available Mythos-class model.
It ships with new safety classifiers that may flag certain prompts in dual-use domains like cyber and bio.
We've added fallbacks: a refused request retries on Claude Opus 4.8 instead of dead-ending.
1
1
6
1,065
📢 you have to pass a new arg to tell the API which model to fallback to!
Don’t just update your model ID to fable, also pass
fallbacks: [{ model: ‘claude-opus-4-8’ }]
Jun 9
Claude Fable 5 is our first generally available Mythos-class model.
It ships with new safety classifiers that may flag certain prompts in dual-use domains like cyber and bio.
We've added fallbacks: a refused request retries on Claude Opus 4.8 instead of dead-ending.
1
1
6
1,065
Lots of people asked how I used Fable to edit its own launch video so I made a video about that!
TLDR it wrote a lot of code & tool calls to use transcription services, ffmpeg, do colorgrading, use the figma mcp, make remotion UI and render it.
I didn't touch a video editor.
292
618
9,211
955,408
The more I work with Claude Managed Agents, the more I fall in love with this pattern:
- Maintain your agent, environment, memory store, vault etc settings with yml and keep them up to date with the ant CLI
- Store IDs for all of the objects in a json file
- Interact with sessions w/ SDKs or wrap the ant CLI (or through the console!)
2
1
16
1,339
The more I work with CMA the more use-cases I start to recognize in the wild. Building a little multiplayer video editor right now that's coming together surprisingly fast.
So much of the _infrastructure_ to bring one of these style apps to production is abstracted away by CMA.
1
3
117