English Français العربية

com.useweckr/weckr

Official registry AI and agents

Ask which of your customers cost more in LLM calls than they pay, per user and per feature.

Source code Website Documentation

Install

Add this to your client's MCP configuration (Claude Desktop: claude_desktop_config.json; Cursor: .cursor/mcp.json; VS Code and others use the same "mcpServers" shape).

NPM package @weckr/mcp
Claude Code 
claude mcp add weckr -e 'WECKR_API_KEY=<your weckr_api_key>' -e 'WECKR_PROJECT_ID=<WECKR_PROJECT_ID>' -e 'WECKR_BASE_URL=<WECKR_BASE_URL>' -- npx -y @weckr/mcp@0.1.2
JSON configuration (Claude Desktop, Cursor, VS Code) 
{
    "mcpServers": {
        "weckr": {
            "command": "npx",
            "args": [
                "-y",
                "@weckr/mcp@0.1.2"
            ],
            "env": {
                "WECKR_API_KEY": "<your weckr_api_key>",
                "WECKR_PROJECT_ID": "<WECKR_PROJECT_ID>",
                "WECKR_BASE_URL": "<WECKR_BASE_URL>"
            }
        }
    }
}

Environment variables

  • WECKR_API_KEY required secret Your Weckr project API key (starts with wk_), from the dashboard sidebar at useweckr.com/dashboard.
  • WECKR_PROJECT_ID Optional. Pins the server to one project id; otherwise it is resolved from the API key.
  • WECKR_BASE_URL Optional. Overrides the API base URL, for self-hosted or staging deployments.

Documentation

From the project's README on GitHub, refreshed weekly. Links open on GitHub.

Contents
  1. Weckr SDKs
  2. Pick your language
  3. Claude Skills
  4. Try it without signing up
  5. Starter templates
  6. Framework guides
  7. How it works
  8. What gets sent
  9. Get an API key
  10. Questions and Support
  11. License

Weckr SDKs

AI cost and margin intelligence for SaaS founders. See exactly which users cost you more than they pay, per LLM call, zero added latency.

Drop the SDK into your app, get a dashboard that shows cost per user / feature / model and recommends cheaper swaps. Set per-plan spending caps the SDK enforces before the LLM call.

Pick your language

PackageInstallSource
TypeScript / Node@weckr/sdknpm install @weckr/sdktypescript/
Pythonweckr-sdkpip install weckr-sdkpython/
Claude / Cursor (MCP)@weckr/mcpnpx -y @weckr/mcpmcp/
Claude Code (Skills)weckr plugin/plugin install weckr@weckrskills/

Each subfolder has full setup docs, examples, and supported-model lists.

Claude Skills

This repo is also an installable Claude Code plugin that bundles four skills, so your AI assistant knows real model prices and the exact Weckr syntax instead of inventing code:

/plugin marketplace add Ghiles3232/weckr-sdks
/plugin install weckr@weckr
SkillWhat Claude learns
weckr-integrationWire the SDK in with the correct two line pattern, errors, and provider notes
weckr-model-pricingCurrent per token prices for OpenAI, Anthropic, Gemini, and Kimi
weckr-cost-estimatorProject what an AI feature costs per call, per user, per month before shipping
weckr-margin-auditFlag which pricing plans go underwater once LLM cost is counted

The pricing skills read the live feed at useweckr.com/pricing.json first (generated from the same table Weckr bills with), and a weekly pricing watcher opens a reviewed PR whenever a provider changes rates, so the numbers track reality instead of drifting. Full explainer at useweckr.com/skills.

Try it without signing up

**Live demo dashboard →**

Click around with seeded data for a fictional SaaS. No signup required.

Starter templates

Clone a full AI SaaS with Weckr already wired in and deploy in minutes. Both are MIT licensed.

  • weckr-nextjs-starter: Next.js 14, Supabase auth, Stripe billing, an OpenAI endpoint, and Weckr tracking cost and margin per user.
  • weckr-fastapi-starter: FastAPI, a Claude endpoint, SQLite, and Weckr tracking cost and margin per user.

Framework guides

Step by step integrations for the stacks people actually ship on, each with the exact code and where the wrap goes:

  • Next.js: route handler pattern, flush() for serverless, streaming notes.
  • FastAPI: endpoint pattern with the authenticated user, zero added dependencies.
  • LangChain: wrap the client at the boundary, not the callbacks.
  • CrewAI: price every agent turn, catch the stuck crew.

How it works

  1. You wrap your LLM calls with wk.chat(client, opts).
  2. The SDK forwards the call to OpenAI / Anthropic / Gemini and returns the original result, with no added latency on the request path.
  3. After it resolves, the SDK fire-and-forgets a log to the Weckr API with (userId, feature, model, tokens, latency, plan).
  4. The backend computes cost server-side from public per-token pricing (clients can't forge cost) and stores (cost, revenue, margin) per request.
  5. The dashboard rolls it up, per user, per feature, per model, and surfaces unprofitable users + cheaper-model recommendations.

What gets sent

Only the call metadata above: model name, token counts, latency, plan, and your userId string. No prompt text or completion text is ever sent.

Get an API key

app.useweckr.com. Free tier, no credit card.

Questions and Support

Have a question about integration or a feature request? Open a discussion: https://github.com/Ghiles3232/weckr-sdks/discussions

We respond to every question.

License

MIT. See LICENSE.

Frequently asked questions

What is com.useweckr/weckr?

Ask which of your customers cost more in LLM calls than they pay, per user and per feature.

How do I install com.useweckr/weckr?

It runs on your machine. In Claude Code: claude mcp add weckr -e 'WECKR_API_KEY=<your weckr_api_key>' -e 'WECKR_PROJECT_ID=<WECKR_PROJECT_ID>' -e 'WECKR_BASE_URL=<WECKR_BASE_URL>' -- npx -y @weckr/mcp@0.1.2. For other clients, paste the JSON configuration from this page into their MCP settings.

Does com.useweckr/weckr need an API key?

Yes. It expects WECKR_API_KEY, which you set in its configuration.

Is com.useweckr/weckr open source?

Yes. Its code is on GitHub under the MIT license.

Who maintains com.useweckr/weckr?

It is published by Ghiles3232. AI.tn lists it but does not maintain or endorse it.

From the official MCP Registry. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.