processtheworldSAFE
forks.world: a library of swim-lane process maps. Create them with Claude or ChatGPT (MCP server, skill, HTTP API).
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
A system that learns how to understand processes.
People explain, demonstrate, record or upload everyday processes. AI turns that input into structured, drillable process maps. A community validates, corrects and improves those maps, and every interaction becomes training data for getting better at understanding processes.
Vision: once the learning loop is proven, apply it to the enterprise: AI interviews employees, observes how work is actually done, and keeps a living representation of how the organization operates. The processes change. The intelligence doesn't.
The app
A Next.js app (same stack as WaitingForPower: Next.js, Prisma, Postgres, Resend, Vercel). The mapper UI lives in src/studio/studio.js; the server owns the engine prompts (src/lib/engine/prompts.ts) and calls Claude (ANTHROPIC_MODEL, default Haiku 5.5).
- Login like WaitingForPower: every browser gets a random key and an anonymous handle (e.g. "BoldPathfinder42"). Adding an email (magic link) keeps maps across devices; an emailed sign-in link restores them on a new device and moves anything made there anonymously into the account. Hashed IPs are used only for rate limits.
- Private, shared, public. Share a process by username or email ("Can edit" / "Can view"); email invites sign people up when accepted. Saves carry a revision, so two editors can't overwrite each other.
- Public processes, decided by consensus. Anyone can "Suggest changes" (a draft tied to the version it started from) and submit it with a note. Others compare before and after (differences listed and marked on the map) and vote "Before is better" / "After is better". A change goes live when "after" leads by 2 (the author counts as one); it's turned down when "before" leads by 2 (
src/lib/consensus.ts). Every version is kept; every vote is stored as a preference pair for training. - Usernames: explorer-themed handles by default, changeable in the account menu, unique ignoring case.
d8150cd8cd0dOBSERVED · 2026-10-11Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add processtheworld --env ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY} --env RESEND_API_KEY=${RESEND_API_KEY} -- npx -y [email protected]{
"mcpServers": {
"processtheworld": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
"RESEND_API_KEY": "${RESEND_API_KEY}"
}
}
}
}Exposed tools (5)
3 read · 2 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
add_comment | write | Comment on a published process, optionally pinned to a step. Like save_process, it signs you up as a guest if you pass no key. |
forks.world | read | Mapping how the world works. A library of process maps, made with AI. |
get_document | read | Get a process as a document: md (Markdown with diagrams and full data), html (the swim-lane flow as a self-contained page), skill (a Claude SKILL.md), or gpt (ChatGPT instructions). Use public_id for a published process, or process_id plus your key for your own. |
list_public | read | List published processes (title, author, tags, steps, likes, id). Search with q, or filter by one tag. Look here FIRST: if a process you need already exists, link it into yours (a step |
save_process | write | Create or update a swim-lane process map and optionally publish it. No account needed: if you don |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (2 observation(s))
- Network
- declared (5 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (6)
process-the-world-skill.zip
import Studio from "../../Studio";
import Studio from "../../Studio";
@vercel/analytics, @types/react-dom
- **Admin dashboard (`/admin`):** only for the verified email [email protected] (fixed in code). Overview metrics, feedback inbox, every session with its transcript, AI changes and actions, pro
Gates applied: no_behavioural_pass, no_license.
d8150cd8cd0dfull audit observations/trust-audit/mcp-server/briandgoldberg__processtheworld.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-11 | d8150cd8cd0d | SAFE | B | 89 | first audit |
Questions
What is the processtheworld MCP server?
forks.world: a library of swim-lane process maps. Create them with Claude or ChatGPT (MCP server, skill, HTTP API).
What tools does processtheworld expose?
5 in total: 3 read-only, 2 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is processtheworld safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does processtheworld need?
It reads ANTHROPIC_API_KEY and RESEND_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How current is this page?
The grade is for one exact copy of the source (d8150cd8cd0d), read on 2026-10-11. The repository is watched and re-audited when it changes.