Reprise MCP server
Which MCP servers are Reprise's own, from the official registry and from asking its own domain, whether they answer, and what they offer. A program checked each of these, nobody copied them from a product page.
read 2026-10-26
- Registry namedesign.reprise/repriseverified in the MCP registry
The registry name sits under reprise.design. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.
- Stateanswerschecked 2026-10-26
- Tools3
- Server calls itselfReprise — un-flatten any flat AI design (bit-perfect)
- Server version1.28.0
- Protocol version2025-06-18
On 2026-10-26 the server answered initialize and tools/list without a sign-in. First answer in 238 ms.
- Whether its tools do the product's job has not been judged yet.
- Checked on 3 of the 7 days it takes to judge reliability.
Said by rules from the measurements above. The likely first call is a model's pick from the tool list; read its description before relying on it.
- INITE MCP scoreafter a weeka score needs 7 daily checks: 3 so far
- Answeredmeasuring3 of the 7 daily checks it takes to print a share
- Median first answer245 ms
- Space in an agent's context≈ 307 tokensthe definitions of all its tools (3), which an agent carries on every turn
- Tool definitionsB · 83 of 100
- Typed results0 of 3 toolsdeclared outputSchema: results come typed, read once
- Not saying whether it only reads or can destroy: 3 of 3
What an agent can do with it
- not sure · Generate colorway and copy variants of an existing ad
- not sure · Swap subject, words and products while keeping the layout
- not sure · Export a finished design as layered SVG or PNG files
Each of the product's use cases, put to Jev against this server's tool list: can an agent do it with these tools?
Worth a look
None of the five warning signs in these definitions.
A program checks the server once a day: initialize and tools/list, no model. Tokens are the definitions' length at about four characters a token. The grade averages six checks of every tool's name, description, schema and annotations; it says how well the tools are described, not whether they work. The score weighs what was measured and nothing else: answering (25%), doing the product's job (25%), definitions (20%), context (10%), use (10%) and how readily a client connects (10%). A part that was not measured is left out, and a hidden instruction caps the score at 40. The state of vendors' MCP servers: method and numbers →
reproduceUn-flatten a flat design image into editable layers and reproduce it bit-perfect, returning the fidelity score. Returns: A JSON string with a bit_perfect flag and fidelity metrics (mae, psnr, exact-match %, stray px).
autodetectAuto-detect the elements (subjects and text) in a design image as relative bounding boxes. Returns: A JSON string with the detected elements (relative bboxes) and their count.
diagnoseScore how faithfully a reproduction matches an original (MAE / PSNR / exact-match % / stray px). Returns: A JSON string with the fidelity metrics.
How to connect
Add the server to your MCP client config. Claude Desktop, Cursor and most other clients read this shape:
{
"mcpServers": {
"reprise": {
"type": "http",
"url": "https://tepesama-reprise-mcp.hf.space/gradio_api/mcp/http"
}
}
}Add it
Claude Code, in a terminal
claude mcp add --transport http reprise https://tepesama-reprise-mcp.hf.space/gradio_api/mcp/httpclaude.ai, ChatGPT and other clients: add https://tepesama-reprise-mcp.hf.space/gradio_api/mcp/http as a custom connector in the settings.
Each line is something a program found when it went looking: the registry entry, the server's own answer, a spec that parses. A missing line means we did not find it, not that it does not exist.
Cookies
Before we count this visit
Analytics and ad measurement tell us which pages and which of our ads do their job: Google Analytics and Google Ads, the Meta pixel, Yandex Metrica, and one first-party id of our own. Nothing is stored until you answer, and declining changes nothing about the site.