A self-hosted workspace for prompts you want to use again. Keep them organised, refine the wording, and test how they perform before your next edit.
Get started · Connect to ChatGPT · Contribute
Find prompts with search, folders and favourites. Write reusable sections with
{{variables}}, fill them in, and copy the result into any chat app. Saved
versions let you compare edits and restore an earlier approach.
Related MCP server: MCP Prompt Manager
Choose Paste a prompt in the library or editor. AI turns unstructured text into XML-style sections and identifies variables with their original values filled in. Review and edit the variables, keep an input fixed, and copy either the template or filled prompt. Apply to the draft with Undo; Save keeps it. Example inputs are separate from template content and require an explicit choice.
Answer clarifying questions, then review a proposed draft section by section. See the wording changes before applying them; undo or save when you're ready.
Give a saved prompt test inputs and choose automatic checks, AI review or your own rating. Inspect model outputs and feedback, then compare runs to understand the effect of an edit. Keep your test cases reusable with CSV import and export.
Screenshots use synthetic demo data, including illustrative AI outputs and scores.
Prompt management works without an AI connection. Prompt conversion, refinement, Try once and evaluation use your configured provider: OpenAI, Anthropic, Gemini, Ollama, vLLM or another OpenAI-compatible endpoint. Hosted calls may incur provider charges; operators can set monthly spending limits.
You'll need Git, Docker with Compose, Make and OpenSSL.
git clone https://github.com/chrysogonus/prompt-maker-studio.git
cd prompt-maker-studio
make setup
make upOpen localhost:3000 and create your account. The first account can register; further registration is closed by default. For AI features, connect a provider under Settings → AI connection, or use the operator's optional server default when configured.
make setup creates local configuration with a generated signing key.
Browser API calls stay on the same origin through the frontend proxy.
Run make down to stop the app. See the Docker guide for
configuration and local email testing.
flowchart LR
Browser[Browser] --> Edge["Caddy · HTTPS"]
MCP["ChatGPT / MCP client"] --> Edge
Edge -->|Web app| UI["Next.js · React"]
Edge -->|REST / MCP / OAuth| API[FastAPI]
API --> DB[(SQLite)]
API -.->|Import · refine · evaluate| AI["Configured AI provider"]Caddy handles production HTTPS and routing. Locally, Next.js proxies /api
to FastAPI. SQLite stores accounts, prompts, versions and evaluation history;
REST and MCP share the same prompt services. AI calls are optional.
Documentation index — current guides, architecture, product records and release history.
Development — local setup, commands and tests.
Deployment — hosting, HTTPS and backups.
Monitoring — dashboards, metrics, alerts and operator logs.
Configuration template — available settings and defaults.
API reference — REST routes; interactive schemas at localhost:8000/docs.
MCP integration — use your library from ChatGPT and other clients.
Early-stage, 0.x. APIs and database schemas may change between releases.
Back up before upgrading; see the changelog. The
0.5.0 release record records publication, immutable
image digests and upgrade checks. Rehearse the migrations on an isolated
production-backup copy before upgrading a live instance. The
0.5.1 release preparation describes the unpublished
candidate currently on dev: dark-mode/editor polish and accessibility fixes.
The candidate introduces no database migrations or REST/MCP contract changes.
Apache 2.0. See NOTICE and third-party notices for attribution.
This server cannot be deployed
Self-hosted AI prompt library: prompts, collections, tags, teams, chains. 29 MCP tools for agents.
Generate contextual prompts and reusable agent skills, evaluate prompts with the 16-dimension Prompt Score, and manage saved work in PromptDrive. Twelve MCP tools also provide authorized access to private Memory for source-grounded answers. Connect over Streamable HTTP using OAuth 2.1 and PKCE. Generation consumes account quota and automatically saves successful results; Memory access follows account permissions and plan limits.
Professional prompt library over remote MCP: 13 verticals, free discovery scope.
Manage, version, and publish LLM prompts with blocks, variables, and evaluations.
