Comparison
Chiltepin vs Archify: who places the shapes
Archify is an open-source agent skill that turns a request into one of five diagram types — architecture, workflow, sequence, dataflow, lifecycle — by having the model write JSON, coordinates included, and checking the result with a geometry validator. Its interactive viewer, with export, a present mode and themes, is polished and ahead of anything Chiltepin renders. Chiltepin makes the opposite bet on the one question that matters: the model writes data, and the renderer places every shape. The head-to-head below is what that bet costs and buys.
Side by side
| Archify | Chiltepin | |
|---|---|---|
| Source format | JSON per diagram, written by the model | Markdown with typed YAML blocks — a document, not a diagram file |
| Who draws the layout | The model: it places coordinates, and a geometry validator catches overlaps and misses | The renderer: the source has no coordinates, so there is no geometry to validate |
| AI agent workflow | An agent skill (MIT); the agent writes JSON, runs the validator, repairs, repeats | An agent skill (MIT); the agent writes YAML, runs chiltepin check, repairs — 8/8 first drafts passed in the head-to-head |
| Validation in CI | The geometry validator can run in a shell; check the repo for CI guidance | chiltepin check — schema, refs, ids, density; exit codes for CI |
| Diagram types | Five: architecture, workflow, sequence, dataflow, lifecycle | Dozens of block types across thirteen families, diagrams and non-diagrams alike |
| Output formats | An interactive HTML viewer with export, present mode and themes — ahead of ours | Static HTML, site, slide deck, PDF; one look, zero runtime JS |
| Hosting and price | Free; runs wherever the agent runs | Free; runs wherever the agent runs, plus a hosted Studio with no signup |
| Open source / licence | MIT | MIT |
| Best for | One striking diagram from an agent, with a viewer to present it in | A whole document from an agent — several blocks, one grammar, checked like code |
The head-to-head
Same model, 13 September 2026, eight requests both tools can express, one run each. The figures are means over the eight; the harness is in the product repository under evals/.
| Archify | Chiltepin | |
|---|---|---|
| Tokens per request | 65K | 57K |
| Wall-clock time | 133 s | 105 s |
| First draft passes its own validator | 1 of 8 | 8 of 8 |
| What one request produces | One diagram | A ~7-block document |
| HTML size per output | 814 KB | 181 KB |
The first-draft gap is the coordinate decision showing up as a number: a model asked to place shapes usually places two of them on top of each other, and the repair loop costs tokens and seconds. The size gap is the viewer: Archify ships an interactive app with every diagram, which is also why its viewer is better. Archify’s export, present mode and themes are ahead of Chiltepin, plainly.
When to pick Archify
Pick Archify if you want one diagram from your agent and a viewer to show it in. Its five types cover the common asks, the geometry validator makes the repair loop converge, and the present mode, export and themes are ready for a meeting today. If the deliverable is the picture rather than the document around it, Archify gets you there with less to learn.
When to pick Chiltepin
Pick Chiltepin if the agent is writing the document, not the picture: the sequence next to the endpoint list next to the decision record, in Markdown, in the repository, checked on every pull request. In the head-to-head it used fewer tokens, finished faster, passed its validator on the first draft eight times out of eight, and produced a whole document at a fifth of the HTML size. What it does not give you yet is Archify’s viewer.
What a Chiltepin block looks like
This is the sequence starter template — one of the diagram types both tools express. Note what is absent: no x, no y, no widths. Actors and messages only.
```sequence
id: seq-example
actors:
- { id: Client, name: Client }
- { id: Server, name: Server }
messages:
- Client -> +Server: request
- alt: cache hit
- Server --> -Client: 200 cached
- else: miss
- Server --> -Client: 200 fresh
- end
```Every block on this site is rendered this way, at build time, by the same pipeline chiltepin build runs. See the block catalog, the 40 documents an agent wrote, or the other comparison, Chiltepin vs Mermaid.
Frequently asked questions
- What is Archify?
- Archify is an open-source (MIT) skill for coding agents that renders five diagram types — architecture, workflow, sequence, dataflow and lifecycle — from JSON the model writes, including the coordinates of each element. It ships a geometry validator that catches overlaps and off-canvas shapes, and an interactive HTML viewer with export, a present mode and themes.
- What is the main design difference?
- Who places the shapes. In Archify the model does, and a validator checks the geometry afterwards. In Chiltepin the model never places anything: it writes data — nodes, edges, lanes, messages — and the renderer computes the layout. That is why Chiltepin has no geometry validator: there is no geometry in the source to get wrong.
- How was the head-to-head measured?
- Same model, same day (13 September 2026), eight requests both tools can express, each run once from a fresh session with only the respective skill installed. We recorded tokens, wall-clock seconds, whether the first draft passed the tool’s own validator, what the output was, and the size of the HTML produced. The numbers are on this page; the eval harness is in the product repository.
- Where is Archify ahead?
- The viewer. Archify’s interactive HTML has export, a present mode and themes, and that polish is ahead of Chiltepin’s static output today. If the deliverable is a single diagram someone will present from a browser, Archify’s viewer is the nicer place to end up.
- Can I use both?
- Yes; they are both skills and do not conflict. A reasonable split is Archify for a one-off diagram you want to present interactively and Chiltepin for the documents that live in the repository and go through review and CI.