Same problem, different audience: humans-at-keyboards vs agents-in-workflows.
PromptPerfect (Jina AI) rewrites prompts for better quality through a web UI — you paste prompt + target model, get optimized version, copy back to your tool. It's polished for solo prompt-engineering work. Promptibus MCP solves the same problem at the agent layer: your AI agent (Claude Desktop / Cursor / Continue.dev) calls `optimize_prompt` programmatically while it's working on a task, with no human copy-paste step.
Use PromptPerfect for hand-tuning your own prompts in their web UI. Use Promptibus when your agent should auto-optimize prompts mid-workflow.
| Feature | Promptibus | PromptPerfect |
|---|---|---|
| Interface | MCP tool calls | Web UI (web app) |
| Use case | Agent calls during reasoning | Human pastes prompt, copies result |
| Multi-model coverage | 80+ models (image, video, text, code, audio) | Mostly text models |
| Cost projection | Built-in (`get_pricing` per model) | No (assumes you know costs) |
| Pre-flight linting | Yes (`lint_prompt` catches broken syntax) | No (output is just a rewrite) |
| Savings reporting | Returns `saved_tokens` + `cost_saved_usd` | No structured savings data |
| Free tier | 100 calls/day, no credit card | Limited free tier, paid for higher use |
You can use both. PromptPerfect for human-in-the-loop tuning during prompt design; Promptibus MCP for agent-in-the-loop optimization at runtime.
Free tier: 100 calls/day. No credit card.