Bridge your AI thinking into your coding workflow.
Meter is a pay-per-thought AI workspace where you chat with every frontier model, run multi-model debates, log structured decisions, and generate architectural blueprints — all on a single postpaid tab.
This repo documents the Meter MCP server — a hosted HTTPS endpoint that connects your IDE's coding agent to your Meter thinking.
meter.chat ── think, debate, decide
│
▼
https://meter.chat/api/mcp ── hosted MCP server
▲
│ HTTPS
│
Your Agent ── Cursor / Claude Code / Codex / Windsurf / Lovable / Replit
Think in Meter. Code with context.
Sign up at meter.chat and copy your key from Settings → API.
Claude Code
claude mcp add meter --transport http https://meter.chat/api/mcp -H "Authorization: Bearer your-api-key"Cursor
Open Settings → MCP Servers → Add Server:
{
"mcpServers": {
"meter": {
"url": "https://meter.chat/api/mcp",
"headers": {
"Authorization": "Bearer your-api-key"
}
}
}
}OpenAI Codex
{
"mcpServers": {
"meter": {
"url": "https://meter.chat/api/mcp",
"headers": {
"Authorization": "Bearer your-api-key"
}
}
}
}Windsurf
{
"mcpServers": {
"meter": {
"url": "https://meter.chat/api/mcp",
"headers": {
"Authorization": "Bearer your-api-key"
}
}
}
}Lovable
{
"mcpServers": {
"meter": {
"url": "https://meter.chat/api/mcp",
"headers": {
"Authorization": "Bearer your-api-key"
}
}
}
}Replit
{
"mcpServers": {
"meter": {
"url": "https://meter.chat/api/mcp",
"headers": {
"Authorization": "Bearer your-api-key"
}
}
}
}Antigravity
{
"mcpServers": {
"meter": {
"url": "https://meter.chat/api/mcp",
"headers": {
"Authorization": "Bearer your-api-key"
}
}
}
}Your agent can now pull your Meter thinking. Try:
- "What did I decide about the auth system?"
- "Show me the blueprint for the API architecture"
- "Search my decisions for anything about database choice"
- "List my recent debates"
6 tools — all read-only. The MCP server pulls your thinking context into the IDE.
| Tool | Description |
|---|---|
get_decisions |
List and search your decision log. Filter by workspace or search term. |
get_decision |
Fetch full detail of a single decision — context, options, rationale. |
get_blueprints |
List and search your blueprints (artifacts). Filter by workspace. |
get_blueprint |
Fetch the full markdown content of a blueprint. |
get_debates |
Browse debate summaries with multi-model synthesis. |
search |
Full-text search across decisions, blueprints, and debates. |
Meter is the first pay-per-thought AI. Every frontier model — Claude, GPT, Gemini, Grok, DeepSeek — on a single postpaid tab. No subscriptions. No rate limits.
Three core primitives:
- Pay-per-thought routing — Every model, one bill. Configurable spend limits. Automatic fallback across providers.
- Structured debates — Pit models against each other in a 4-phase adversarial framework. Three models from three labs attack each other's logic.
- Agent Spec Kit — Synthesize decisions and debates into artifacts your coding agent needs (
ARCHITECTURE.md,DECISIONS.md,CLAUDE.md,.cursorrules), committed directly to GitHub.
This MCP server is the bridge. You think in Meter. Your agent codes with that context.
Contributions welcome. See CLAUDE.md for AI agent instructions.
Think in Meter. Code with context.