Open-source prompt operations

Every prompt has a record.

Version, evaluate, and manage production prompts with a local-first ledger built for serious engineering workflows.

Version ledger

support-agent.promptv12 to v13

Current

- Be helpful and concise.- Answer every request.+ Answer using verified context.+ Cite the source record. If context is incomplete:+ Ask one clarifying question. Return: {answer}

Commit7f3a91c

Authorlocal

StatusSaved

SELF-HOSTEDLocal control

SQLITEPortable data

MITOpen source

14MCP tools

4Resources

OpenAI / Anthropic / Gemini / Ollama

01 / Capabilities

A disciplined prompt workbench.

Version control, evaluation, and delivery tools arranged around a local source of truth.

01DIFF

Version Prompts

Immutable version history with diffs and rollback. Track every change to your prompts like code.

02EVAL

Evaluate Everything

Run prompts against datasets with automatic scoring, cost tracking, and metrics for 23+ LLM models.

03MCP

MCP Server

Connect directly to MCP clients with 14 tools and 4 resources for full prompt lifecycle management.

04CLI

CLI & REST API

Use the command line or HTTP API to fit PromptVault into existing workflows and custom tooling.

05LLM

Multi-Provider

OpenAI, Anthropic, Google Gemini, and Ollama. Switch providers with a single config change.

06LOCAL

Local-First

A SQLite database on your machine. No cloud dependency, no vendor lock-in, and an MIT license.

02 / First run

Quick Start

Three commands from local checkout to a repeatable prompt evaluation.

Install

shell

git clone https://github.com/KrishBnsl/promptVault.git
cd promptVault
uv sync

Create Prompt

shell

promptctl prompt create qa-bot \
  --content "Answer: {question}" \
  --tags "qa"

Evaluate

shell

promptctl eval run qa-bot \
  --dataset test-data \
  --model-config '{"provider":"gemini"}'