Limner

Originally built for an HR consultant client developing gamified training programs with VGA-era visual aesthetics. Now being rebuilt on Cloudflare Workers as a public-beta framework.

Limner image asset pipeline
vinsonconsulting/ limner
Private

An image and taste agent built as an MCP server: 15+ namespaced tools for generating, composing, and recalling work, with durable memory in D1 so a session remembers what it has made. The agent's code runs on Claude Managed Agents in Cloudflare's isolated execution environment, and it ships three ways: an OAuth-gated Workers endpoint, a local stdio server, and a Claude Desktop bundle. This is the 'Rasa' variant of the Limner agent, built to be useful out of the box, ready to learn your styles and flows and then hold you to them, with strong opinions about the work it produces.

MCP Cloudflare Workers Claude Managed Agents D1 OpenAI gpt-image-1 Recraft

Context

In early 2025, an HR consulting client reached out to me for help on a project that ‘required’ training materials with VGA-era video game aesthetics. The ask was for hundreds of stylistically consistent pixel-art assets for gamified onboarding modules, where the visual language was attempting to do curricular work. Off-the-shelf image generators couldn’t hit the aesthetic reliably, and human art direction at that volume was the cost we were trying to reduce.

The first version (Limner: Pixel) shipped that engagement. The framework that came out of it became a family.

Initial Approach

A pipeline-plus-judgment architecture:

  1. MidJourney for base imagery generation, effectively the editorial layer.
  2. PixelLab (via a custom Limner agent) for pixel-art translation
  3. Automated quality gate. The agent evaluates each output against stylistic criteria (palette adherence, dithering patterns, character density, VGA-era conventions) and rejects non-conforming results before they reach the output queue
  4. Vectorization for final output

The quality gate is the critical piece. Without it, this is a chained tool pipeline. With it, the agent is making aesthetic judgments at each stage and refusing to pass through work that fails those judgments.

Outcome

Multiple batches of 500+ uniformly stylized assets, produced with zero human intervention once the quality gate was tuned. The HR consultant shipped a demo of the gamified onboarding modules. The framework, the discipline of building an agent that throws away its own work when it doesn’t meet a bar, generalized into a family.

The Limner family

Limner: Pixel (production, public-beta rebuild in progress). The original framework. Currently being rebuilt on Cloudflare Workers as the “CF Limner” public-beta release on top of Rasa, the new substrate base. A prospect pipeline is gated on the release.

Limner: ASCII (in development). Self-initiated extension of the framework. ASCII / BBS-era aesthetics, built on Rasa.

Rasa (new base, in development). A medium-agnostic Limner substrate that the Pixel rebuild, Limner: ASCII, and future variants all run on. Same quality-gating discipline, multiple aesthetic targets.