MCP Tasks Let Plugins Run Long Background Jobs
A new Tasks primitive for MCP lets Claude hand off long-running work to plugins, like deep research, and get a callback when the job finishes.
ClaudeDevs·Fri, Oct 23 2:24pm ET
Elicitations Let MCP Plugins Request Mid-Task User Input
Elicitations are a new MCP primitive that lets plugins pause and ask users for input during a running task, like confirming a refund or supplying a missing detail.
ClaudeDevs·Fri, Oct 23 2:24pm ET
MCP Tasks and Elicitations Ship in Claude Code CLI Today
Tasks and Elicitations are available now in the Claude Code CLI, with a rollout to more Claude products coming. Plugin developers can add both primitives to their MCP servers today.
ClaudeDevs·Fri, Oct 23 2:24pm ET
Claude on Google Cloud Generally Available With Full API Access
Claude Models on Google Cloud (Anthropic managed) hit general availability, bringing the full Claude Platform feature set including Managed Agents, code execution, and Opus 5.5 fast mode to GCP accounts with native IAM billing and commitment retirement.
ClaudeDevs·Thu, Oct 22 5:27pm ET
Anthropic Investigating Claude Code Usage Limit Spike
Multiple users flagged that Claude Code usage is draining dramatically faster than before, with some reporting 5x higher consumption. Boris confirmed the team is actively investigating and will report back soon.
bcherny·Thu, Oct 22 7:48pm ET
Lydia Asks for Session Details on Usage Limit Bug
A user flagged two specific patterns: context compression consuming more usage than expected, and limits behaving strangely in long conversations. Lydia asked follow-up questions about cold prompt cache behavior, subagent use, model and effort level, and requested /feedback IDs to dig in.
lydiahallie·Thu, Oct 22 6:41pm ET
More Stories
Boris Asks Users to Run Diagnostic Command for Usage Issues
Multiple users reported hitting weekly usage limits faster than expected. Boris shared a shell command to list recent Claude project session files, likely to help diagnose whether local session bloat is contributing to the drain.
bcherny·Thu, Oct 22 8:59pm ET
Bcherny DMs User to Debug Token Limit Issue
A team member reached out directly to a user who reported an ongoing issue with hitting usage limits faster than expected, to gather more debugging information.
bcherny·Fri, Oct 23 12:02pm ET
Boris Contacts User With Max Limit Drain Issue
A user reported their account fully exhausted its weekly budget three days after a Monday reset, without running heavy workloads. Boris thanked them and DMed for more details.
bcherny·Thu, Oct 22 7:40pm ET
Boris DMs User Who Burned 87% in 36 Hours
A user burned through 87% of their freshly reset weekly budget in 36 hours, calling it unprecedented. Boris DMed for more information.
bcherny·Thu, Oct 22 7:40pm ET
Boris DMs Users Reporting Claude Code Limit Drain
A user reported burning through two full 20x plan resets in a single week. Boris DMed for more details as the team investigates usage spikes.
bcherny·Thu, Oct 22 8:00pm ET
Boris Collecting DMs on Credit Drain Reports
Another user reported credits depleting noticeably faster. Boris messaged directly for more information as the team investigates.
bcherny·Thu, Oct 22 7:59pm ET
User Tries and Fails to Burn Claude Max Limits
A user ran 30-plus agents across three projects, generated videos, and built features in a deliberate attempt to exhaust their Claude Max budget before a free reset expired, and still couldn't drain it. Boris replied with a heart.
bcherny·Thu, Oct 22 7:10pm ET
Bun AOT Merge Next Week, Claude Code Test to Follow
The fourth iteration of Bun's AOT compiler is nearly ready on performance and will merge next week, followed by testing on Claude Code. The compiler targets standard TypeScript with no special syntax, dropping earlier requirements for `as` casts or unusual argument handling.
jarredsumner·Fri, Oct 23 10:06am ET