# Promptic

> An AI optimization platform for LLM tracing, prompt tuning, agent benchmarking, and model comparison using OpenTelemetry-native instrumentation.

Promptic is a GenAI optimization platform built for developers and teams who need to systematically improve the quality, cost, and latency of LLM-powered applications. It combines OpenTelemetry-native tracing with automated prompt optimization, agent benchmarking, and model comparison — all tied to user-defined business metrics and real production data.

## What It Is

Promptic sits at the intersection of observability and optimization for AI applications. Rather than treating tracing as an end goal, the platform uses captured LLM call data as the starting point for iterative improvement: upload your use-case data, define KPIs, and let Promptic benchmark and tune every layer of your GenAI stack automatically. It supports prompt optimization, agent architecture comparison, MCP tool selection tuning, and information extraction workflows.

## How the Optimization Loop Works

The core workflow follows four steps:
- **Upload data** — bring real inputs and expected outputs as a fixed evaluation dataset
- **Set KPIs** — choose quality, cost, and latency targets
- **Surface insights** — Promptic identifies where each candidate wins or breaks down and the patterns behind every score
- **Optimize automatically** — the platform tunes prompts, models, and tools against your data, then ships the best-value configuration

The homepage illustrates a concrete example: a customer support intent classifier starting at 54.8% accuracy reaches 97.01% after three optimizer iterations, while also reducing prompt token usage.

## OpenTelemetry-Native Tracing

Promptic captures every LLM call, tool call, and agent step — tokens, cost, and latency per span — using OpenTelemetry as the underlying standard. The trace view shows individual spans (plan, search, rerank, reason, respond) with model, provider, duration, token counts, cost, and input/output for each. The platform positions tracing as the "doorway" to optimization rather than the destination, enabling teams to turn real agent behavior into actionable improvement data.

## Coding Agent Integration

Promptic exposes an MCP server and SDK so coding agents (such as Claude Code) can drive the optimization loop autonomously. The documented workflow shows an agent reading the current setup, defining candidate architectures, pinning evaluation datasets and KPIs, benchmarking variants head-to-head, inspecting failing traces, implementing the winning architecture, validating it against the same dataset, and shipping the verified change — all through Promptic tool calls.

## Platform Architecture and Integrations

The platform is organized around Organizations, AI Applications, and AI Components. It supports:
- **Model providers**: OpenAI, Anthropic, Mistral, Google Gemini, xAI Grok, and others
- **Access modes**: bring-your-own model keys (BYOK) or Promptic-managed model billing
- **Interfaces**: dashboard, Python SDK, CLI, and REST API (OpenAPI contract)
- **Deployment**: cloud (EU data region by default), Private Cloud, and BYOC for enterprise
- **Governance**: RBAC, SSO/SAML, SCIM, audit logs, and CI/CD gates on higher tiers

## Current Status

Promptic is actively available with a live dashboard, public documentation, SDK quickstart, and a Product Hunt listing. The pricing page references 2026 copyright, and the platform UI shows model names such as GPT-5.4 Mini, GPT-5.5, Claude Opus 4.8, and Gemini 3.1 Pro, indicating the product is tracking current frontier model releases. An enterprise program is available for organizations needing custom deployment, committed usage, and dedicated support.

## Features
- OpenTelemetry-native LLM tracing
- Automated prompt optimization
- Agent architecture benchmarking
- Model comparison (quality, cost, latency)
- MCP tool selection optimization
- Information extraction optimization
- Coding agent integration via MCP server
- Bring your own model keys (BYOK)
- Promptic-managed model billing
- Python SDK and REST API
- CI/CD gates
- RBAC and access governance
- SSO/SAML, SCIM, audit logs (Enterprise)
- EU data residency
- Private Cloud / BYOC deployment
- Dataset and test case management
- AI-assisted data and judge generation
- Deployment of optimized prompts to production

## Integrations
OpenAI, Anthropic, Google Gemini, Mistral, xAI Grok, OpenTelemetry, Python SDK, REST API, MCP (Model Context Protocol), CI/CD pipelines

## Platforms
WEB, API, CLI, DEVELOPER_SDK

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://promptic.eu
- Documentation: https://promptic.eu/docs
- EveryDev.ai: https://www.everydev.ai/tools/promptic