openflowkit
llms.txt923 starsChanged 25 days agoMITView on GitHub
Diagrams agents write, kept valid, editable and saved next to the code. Install as a skill: skills/openflowkit/SKILL.md in https://github.com/openflowkit/openflowkit Write Mermaid or OpenFlow DSL, hand it to the openflowkit MCP server, and save the result as a .openflow.json beside the code it explains: it opens at app.openflowkit.com (Open file) and reads back. When to draw: a flow with branches, a system with more than three moving parts, a protocol, a data model, a state machine. Not a list, a single…
We split this llms.txt into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 6
Write Mermaid or OpenFlow DSL, hand it to the openflowkit MCP server, and save the result as a .openflow.json beside the code it explains: …
- Distinctive: line 19
Write the diagram in Mermaid (what you already know) or OpenFlow DSL (below).
- Very specific: line 116
Do not place coordinates; layout is computed.
- Very specific: line 117
Check before you claim: validate_openflow_dsl passes, then save, then report the path.
≈ “Common” is an estimate: at least 1 in 5 of the 3,225 llms.txt files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
# OpenFlowKit
> Diagrams agents write, kept valid, editable and saved next to the code.
> Install as a skill: skills/openflowkit/SKILL.md in https://github.com/openflowkit/openflowkit
Write Mermaid or OpenFlow DSL, hand it to the `openflowkit` MCP server, and save the result as a
`.openflow.json` beside the code it explains: it opens at app.openflowkit.com (Open file) and reads back.
**When to draw:** a flow with branches, a system with more than three moving parts, a protocol,
a data model, a state machine. Not a list, a single call, or what one sentence explains.
## Setup (once)
`claude mcp add -s user openflowkit -- npx -y openflowkit mcp` (any MCP client: `npx -y openflowkit mcp`); it
calls no outside service. In claude.ai or ChatGPT, `https://mcp.openflowkit.com/mcp` draws chat cards only.
## Draw and save
1. Write the diagram in Mermaid (what you already know) or OpenFlow DSL (below).
2. `validate_openflow_dsl` with `{ dsl }` until `ok`. Fix every `error`: a line the parser dropped is one,
with its line and a hint. Warnings are guesses; read them. For Mermaid (or Structurizr DSL, D2),
`converted.losses` name what did not carry over and `converted.dsl` is what it became.
3. With an editor paired (`whoami`; if not, ask the user to open app.openflowkit.com/#/connect),
`create_diagram` with `{ dsl }` draws where the user is looking. Otherwise use file mode:
`openflow_create`, then `create_diagram` with `{ dsl, documentId }`.
4. `openflow_save` with `{ path }` (plus `documentId` in file mode) beside the code, e.g.
`src/payments/checkout-flow.openflow.json`; `svg: true` also writes a `.svg` (for a README).
5. Tell the user the path, and what you left out.
## Change a diagram
`get_diagram` returns the DSL and `frameId` (file mode: `openflow_open` first, then pass
`documentId`) → edit the text → `update_diagram` with `{ frameId, dsl }` → `openflow_save`.
Not `create_diagram` again: that draws a second diagram. `update_diagram` redraws from the text, so
`canvasEdits` (the user's recolours, moves, renames) are lost unless you merge each `carry` into that
node's one `[ … ]` list; `discarded` lists any lost. "The box I selected" is `selection` in `get_document`.
## Draw this repo's architecture
`discover_architecture` with `{ path }` proposes a C4 workspace (`dsl`) from the code. Review its
names, `create_diagram` with it, and write the text to `architecture.ofk` at the repo root; later,
`drift_report` with `{ path }` says what the code gained or lost since.
## OpenFlow DSL in one screen
First line is the family: `flowchart`, `architecture`, `sequence`, `state`, `erd`, `class`,
`mindmap`, `gitgraph`, optionally with a direction (`flowchart right`). Lines are statements;
`//` starts a comment. Nodes are declared by name; edges declare their ends.
```dsl
flowchart right
title: Checkout
group Payments {
Pay [diamond]
Stripe [icon: tabler/brand-stripe]
}
Cart -> Pay : card ok?
Pay -> Done [ellipse, green] : yes
Pay --> Cart : no [red]
Pay -> Stripe : charge [thick]
```
- Arrows: `->` solid, `-->` dashed, `<->` both ends, `--` no heads, `A -> B : label`.
- Attributes go in `[ … ]`: a shape, a colour, a fill, flags, or `key: value`.
- Shapes: `rect rounded circle ellipse diamond cylinder hexagon cloud doc note parallelogram person queue component browser mobile`
- Colours: `blue green red orange violet teal pink yellow gray` or `#hex`.
- Icons: `icon: aws/lambda`, `gcp/pubsub`, `tech/react`, `tabler/user`. A node whose label names a
technology (`Postgres`, `Redis`, `S3 bucket`) gets its icon by itself; `icon: none` stops it.
`iconpos: left` draws an icon card as a row, the icon beside its name. `search_icons` finds exact ids.
- `group Name { … }` boxes the nodes first mentioned inside it, so open groups before edges.
An edge to `Name` ends on the group itself.
`note Name : text` pins a sticky to a node.
```dsl
sequence
participant Browser [person]
Browser -> API : POST /login
API -> DB : find user
DB --> API : row
alt valid {
API --> Browser : 200 + token
} else invalid {
API --> Browser : 401
}
```
```dsl
erd
users {
id uuid pk
email text unique
}
orders {
id uuid pk
user_id uuid fk
}
users ||--o{ orders : places
```
For any other family, or more detail, call `get_syntax` with the family name.
## Mermaid works too
```mermaid
flowchart LR
A[Client] -->|HTTPS| B(API)
B --> C[(Postgres)]
```
Converts: flowchart, sequenceDiagram, stateDiagram, classDiagram, erDiagram, mindmap, architecture-beta,
gitGraph; not gantt, pie, journey (write `chart bar` DSL, see `get_syntax`). Structurizr and D2 text convert too.
## Rules
- One diagram per idea; split before forty nodes. Name nodes as the code does (`OrderService`).
- Do not place coordinates; layout is computed. Change the text and `update_diagram`.
- Check before you claim: `validate_openflow_dsl` passes, then save, then report the path.
Other versions1 project uses a version of this.
Word for word · 1
- openflowkit/openflowkit★ 909100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository2 other files it gives its agents.
- AGENTS.mdAGENTS.md
- openflowkitSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

