Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Question on how nanoclaw works #1666

etlnerd started this conversation in General
Discussion options

Based on this statement in the readme
"Best harness, best model. NanoClaw runs on the Claude Agent SDK, which means you're running Claude Code directly. Claude Code is highly capable and its coding and problem-solving capabilities allow it to modify and expand NanoClaw and tailor it to each user."

With the recent announcement that Anthropic is killing OpenClaw users, how does Nanoclaw run? I have a max subscription, so can I expect that I would use something like the claude setup-token, or is this purely API token usage based?

You must be logged in to vote

Replies: 1 comment

Comment options

Hey @etlnerd, Gavriel shared a bit more around this on the Community Discord server. Hope it helps!

Update on Anthropic's subscription policy changes

You've probably seen the news: Anthropic is ending subscription coverage for third-party harnesses, starting with OpenClaw. We've seen the questions and anxiety in the community over the past 48 hours, so here's where things stand for NanoClaw.

NanoClaw is built on Agent SDK. Anthropic has indicated that tools built on Agent SDK are in a different category than third-party harnesses that were accessing subscriptions. They've said they're working on providing clearer guidance for this category soon. We don't want to get ahead of that. We're not going to claim anything until we have explicit clarity. But the architectural distinction is real and meaningful.

We'll share updates as soon as Anthropic provides more specific guidance.

We take being good citizens of Anthropic's infrastructure seriously. A few things we've already done and continue to do:

  • Compaction threshold is set to 165K tokens as the default. Sessions don't run to the edge of the context window.
  • No heartbeat polling. NanoClaw doesn't wake up the agent on a regular schedule in the background by default. Only when a message is sent or a task is scheduled.
  • Scheduled tasks can run a detection script before invoking the agent, so we're not spinning up sessions unnecessarily.
  • We'll be doing more to make it transparent where your agent stands in regard to its context window and to give you better controls to compact and clear.

We're watching this closely and we'll keep you posted as things develop.

You must be logged in to vote
0 replies
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
2 participants

AltStyle によって変換されたページ (->オリジナル) /