PromptFlowEngine
Prompt engineering, security and PromptOps for developers and teams.
A deterministic engine that analyses, structures, secures and optimises AI prompts — in the browser, VS Code, AI agents over MCP, CI and an HTTP API.
Status:Available — Website, browser workspace and HTTP API are live. Verified 15 November 2026.
The problem
Developers send prompts to AI tools with missing requirements, wasted tokens and sometimes credentials or personal data. Teams adopting AI coding assistants have no consistent way to check prompt quality or enforce a policy before data leaves the machine.
The solution
PromptFlowEngine turns a rough request plus project context into a structured engineering prompt — objective, requirements, constraints, edge cases, tests and acceptance criteria — and checks every prompt for ambiguity, wasted tokens, prompt-injection patterns, leaked secrets and personal data. One engine runs everywhere, with no model calls and no prompt uploads required.
What it does today
- Prompt analysis, restructuring and quality scoring, with checks that every original requirement is preserved.
- Detection and local redaction of secrets, personal data and prompt-injection patterns.
- Token and cost comparison across major model families, before and after optimisation.
- Browser workspace that runs the full engine in the tab, with no sign-up.
- Public HTTP API that needs no key.
- VS Code extension that turns a change request and the relevant code into a specification-style prompt.
- Chrome extension for pre-submit checks on popular AI chat sites (installed unpacked).
- MCP server with read-only tools for AI agents, and CLI commands for CI gating.
Architecture overview
One pure-TypeScript engine package is embedded in every surface, so a prompt gets the same analysis in the browser, the editor, an agent, CI and the API. The website and API run on Cloudflare’s edge.
Layer 1 · Surfaces
- Web workspace
- VS Code extension
- Chrome extension
- CLI
- MCP server
- HTTP API
Layer 2 · Core
- Deterministic prompt engine (parse, analyse, optimise, evaluate)
- Sensitive-data and injection detection
Layer 3 · Platform
- Cloudflare Workers + static assets
- Cloudflare D1 (counts and categories, never prompt text)
On the roadmap
Not yet available. Plans may change.
- Self-hosted enterprise AI gateway.
- Wider enterprise policy and reporting features.