ANTHROPIC 'CLAUDE CODE' UPDATE: SUB-AGENTS, LSP HOOKS, AND CLAUDE ULTRA MODEL
A recent video reports that Anthropic updated 'Claude Code' with sub-agents for decomposing tasks, integration with Language Server Protocols (LSPs), and a new ...
A recent video reports that Anthropic updated 'Claude Code' with sub-agents for decomposing tasks, integration with Language Server Protocols (LSPs), and a new 'Claude Ultra' coding model. The video does not show official docs, so treat details as preliminary. If accurate, these features aim to improve code navigation and task automation across large repos and multi-language backends.
Sub-agents could break backend changes (APIs, migrations, tests) into smaller, reviewable steps.
LSP integration may anchor suggestions to real symbols and types, reducing hallucinations in large codebases.
-
terminal
Pilot in a monorepo with pyright/gopls and measure suggestion accuracy, latency, and PR rework rates over one sprint.
-
terminal
Prototype a sub-agent flow for schema migration generation and test updates, gated by PR comments only (no direct writes).
Legacy codebase integration strategies...
- 01.
Enable LSP-aware read-only suggestions first and apply changes via PRs to avoid surprises in legacy services.
- 02.
Map sub-agent roles to existing CI steps (lint, tests, migrations) and gate with current approvals and audit logs.
Fresh architecture paradigms...
- 01.
Standardize LSP configs, code owners, and test runners early so agents have consistent boundaries and tools.
- 02.
Define service-scoped agent roles and tool contracts (build, lint, migrations) to keep automation predictable.