Minimal, fast inline autocomplete for any text — code, LaTeX, prose, notes, and inline Q&A — powered by Claude. Completions appear as ghost text as you type; press Tab to accept.
Requirements
This extension calls the Anthropic API using your own key. Set it as an environment variable:
export ANTHROPIC_API_KEY="sk-ant-..."
Important (macOS/Linux): apps launched from the GUI (Dock/app launcher) do not inherit environment variables from your shell profile. Either add the export to your shell profile and launch VS Code from a terminal (code .), or set the key so the VS Code process can see it. The extension will show a warning if it can't find the key on startup.
Your key is read directly from the environment by the Anthropic SDK — it is never stored by the extension or bundled into the package.
Features
Inline completions everywhere — works in any file type, not just code.
Inline Q&A — end a line with a question and the completion is the answer, formatted to match your document.
Auto / Manual modes — toggle whether completions trigger automatically as you type or on demand.
Live model list — pick your model from the list fetched directly from the Anthropic Models API, so it never goes stale.
Cost tracking & monthly budget — per-day/week/month token and cost stats, with an optional monthly spend cap.
Streaming — completions stream in token-by-token.
Commands
Command
Description
Verse: Toggle Mode (Auto/Manual)
Switch between automatic and on-demand completions. Also opens the settings menu, including Select Model.
Verse: Show Usage Stats
View token usage, cost, and budget projection.
Settings
Setting
Default
Description
verse-inline-completions.model
claude-haiku-4-5
Claude model ID. Use the Select Model menu to pick from the live API list, or enter any valid model ID.
verse-inline-completions.maxTokens
1024
Max tokens per completion.
verse-inline-completions.customInstructions
""
Extra system-prompt instructions.
verse-inline-completions.temperature
0
Sampling temperature (-1 to omit).
verse-inline-completions.topP
-1
Nucleus sampling (-1 to omit).
verse-inline-completions.topK
-1
Top-K sampling (-1 to omit).
verse-inline-completions.prefixLines
200
Lines of context before the cursor.
verse-inline-completions.suffixLines
50
Lines of context after the cursor.
verse-inline-completions.debounceMs
800
Delay before an auto-completion fires.
verse-inline-completions.streaming
true
Stream completions token-by-token.
verse-inline-completions.monthlyBudget
0
Monthly spend cap in USD (0 = unlimited).
Notes on cost tracking
The list of available models is pulled live from the Anthropic API, which does not expose pricing. Per-model prices used for cost tracking are maintained locally and sourced from the Anthropic pricing page. A model the API returns that has no local price is still fully usable — its usage is simply excluded from cost tracking.
