v3.1.7: Context Engineering stats, auto-detected model, Generate Glossary→

Count your tokens.
Like beads on an abacus.

Every token counts — monitor budgets, estimate costs, optimize context, and cut AI spending before it adds up. A VS Code extension, MCP server, and CLI that makes the invisible visible.

Learn MoreStar on GitHub
VS CodeVS CodeClaude CodeClaude CodeMCP
500+installsarXiv2604.0829017modelsOpensource
@tokalator
Tokalator token budget dashboard in VS Code — LOW budget, Context Engineering stats, budget breakdown

Features

BPE Tokenization

Real token counts with Claude BPE and OpenAI o200k_base encoders.

v3.1.7

Context Budget Meter

Live sidebar showing token usage vs model context window with Budget Level indicators.

v3.1.7

Tab Relevance Scoring

5-signal scoring: language match, import relationships, path similarity, edit recency, diagnostics.

v3.1.7

Cost-per-Turn Estimate

Know the dollar cost before you send. Per-MTok pricing for 17 models.

v3.1.7

17 Model Profiles

Claude Opus/Sonnet/Haiku 4.x, GPT-5.x, o4-mini, Gemini 3.x/2.5 with live pricing.

v3.1.7

Next Turn Preview

Project total tokens after your next turn to decide whether to send or restart.

v3.1.7

Workspace Security Scan

Detect Sensitive Files (.env, .pem, credentials) and generate deny rules for Claude and Copilot.

ProUpgrade to Pro →

Smart Model Routing

Automatically select the most cost-effective model based on task complexity.

ProUpgrade to Pro →

Team Analytics

Usage dashboards, audit logs, cost insights, and spend controls across your organization.

EnterpriseContact Sales →

Quick Start

  1. 1Install from the VS Code Marketplace or run `ext install vfaraji89.tokalator`
  2. 2Click the Tokalator icon in the Activity Bar — the sidebar dashboard opens
  3. 3Type `@tokalator /count` in Copilot or Claude Chat to see your token budget
  4. 4Pin important files with `/pin` so they always stay in context
  5. 5Run `/optimize` to get a full health report on your context window

Under the hood

Tokenizers
Claude BPE (Anthropic), o200k_base (OpenAI), heuristic (Gemini)
Models
17 profiles — Claude Opus/Sonnet/Haiku 4.x, GPT-5.5, GPT-5.4 family, GPT-5.3 Codex, o4-mini, Gemini 3.x, Gemini 2.5
Caching
Prompt caching (Anthropic 90%), automatic (OpenAI 50%), context caching (Google 75%)
Scoring
Relevance from 0 to 1 based on imports, language, path similarity, recency, and diagnostics
Pricing
Live cost estimation per turn — input/output breakdown with per-MTok rates for all 17 models
MCP & CLI
count_tokens, estimate_budget, preview_turn, list_models — works in Claude Code via stdio MCP or as a standalone terminal CLI

All Commands

Available through @tokalator in VS Code Chat.

/count

Show current token count and budget status

/breakdown

Show where your tokens are going

/optimize

Close low-relevance tabs to free up tokens

/pin

Pin a file so it's always included

/unpin

Unpin a file so it returns to normal relevance scoring

/instructions

List, count, and estimate tokens for instruction files in workspace

/terminology-gen

Scan terminology sources and show token compression potential

/model

Show or switch the active AI model

/compaction

Show per-turn token growth and compaction recommendations

/preview

Preview token cost of your next message before sending

/reset

Reset session state (chat turn counter)

/exit

End the current session and save summary

View on GitHub →

Beyond VS Code

Token budget awareness wherever you work — not just in the editor.

1

MCP Server for Claude Code

Real Claude BPE token counting inside Claude Code via stdio MCP transport. Count tokens, estimate file budgets, preview turn cost — zero API calls.

claude mcp add --transport stdio tokalator -- node tokalator-mcp/build/index.js
MCP
2

Standalone CLI

Terminal-first token budgeting for SSH sessions, containers, and non-VS Code workflows. Run tokalator count, budget, preview, or models anywhere.

npm install -g tokalator
CLI
Pro

Built for teams that ship with AI

The free extension keeps your own context in check. Tokalator Pro takes that across your team, your pipeline, and your sensitive code, so context stays a managed budget instead of a surprise.

See it across the team

Shared context budgets and a team dashboard, so the whole org sees what is filling the window, not just the person who opened the file.

Catch regressions early

Historical token analytics across days, sprints, and models. Spot context spend creeping up long before it shows on the invoice.

Guardrails in CI

Budget gates fail the build when context blows past your threshold, keeping AI agents honest inside the pipeline.

Keep secrets out of context

Secure Workspace shields sensitive files from the model, and custom model profiles plug in your internal or fine-tuned models.

See what is in Pro →

Every token you save
makes the next prompt better.

Open source. Community shaped. Free forever.

Star on GitHub