Curb Cut Mcp

io.github.parthsevak2Published by GitHub account parthsevak2v1.0.0Updated Oct 24, 2026

Workplace accommodation options, cost evidence and drafts. Looks up and drafts; never sends.

Overview

AI-generated overview

Local MCP server for Curb Cut that lets an assistant look up workplace accommodation options and cost evidence and draft requests, without sending anything.

What you can do
  • Look up workplace accommodation options and cost evidence from a Curb Cut deployment
  • Draft accommodation requests in the worker's own words, with the site URL and support address quoted so the person can send them
  • Hand off to a human on the access desk when a request cannot be completed
  • Five tools in total, none of which can send, submit or store a message on the person's behalf
When to use it
Useful when a worker wants to explore accommodation options and cost figures and prepare a request through their own assistant, without disclosing a diagnosis or committing to send anything. Suited to setups already running a Curb Cut Salesforce org with an access desk behind it.
Requirements
Runs locally over stdio via npx (Node.js). Needs a Curb Cut Salesforce org and the Salesforce CLI, with SF_ORG_ALIAS naming the org (defaults to curbcut) and SF_CLI_MODULES if the CLI is installed in an unusual location. CURB_CUT_SITE and CURB_CUT_SUPPORT_EMAIL supply the site URL and support address quoted in drafts; TWILIO_NUMBER is optional. No authentication is declared.
Before you install
Drafts are written for the person to send themselves; the server cannot send, so nothing should be treated as submitted. Accommodation requests concern disability and health, so review what is passed to the assistant and keep the no-diagnosis stance. The server depends on a Salesforce org and CLI credentials configured on the machine.

Installation

In SourceWeft

  1. Open Curb Cut Mcp in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Curb Cut

Curb Cut lets any worker ask for a workplace accommodation in their own words, on the web with no login, by text or a call from a basic phone, by email, in a Slack DM, or through their own assistant, without ever disclosing a diagnosis. An Agentforce agent holds the conversation; deterministic Apex decides what is stored and sent; a person on the access desk answers. First place, Builder Track, Agentforce for Good Hackathon, Dreamforce 2026.

git clone https://github.com/parthsevak2/curb-cutcd curb-cutpython3 tests/invariants.py     # 521 structural checks, offline, about a second

Feel it before you read anything: type one sentence at parthsevak2.github.io/curb-cut/try.html and watch the product's own ranking answer, entirely in your browser. The landing page with the films is parthsevak2.github.io/curb-cut, and a print-ready break-room poster is poster.html.

The story behind it, ten minutes: https://parthsevak2.github.io/curb-cut/story.html

To install it into your own Salesforce org, follow DEPLOY.md. It needs Agentforce enabled; text and voice also need your own Twilio number and a small relay (channels/), and Slack needs the app created in your workspace from the manifest. Four Custom Labels carry your deployment's number, site URL and support address.

Why it exists

The process for requesting a disability accommodation is itself an accessibility barrier.

To get help you must first navigate a system designed by and for people without your disability. Find the policy. Learn its vocabulary. Hold a corporate login. Complete a form. And disclose something you may be afraid to disclose, to a stranger, in writing, in a language that may not be your first.

Cost usually isn't the barrier.

The film, 4 minutes 59 seconds, with every channel shown live: https://youtu.be/yleHLiwWRKA. A 26-second recording of trying the ask page: https://drive.google.com/file/d/13NApu8rq3MpJjuGG2yzKuJ2P02tKWwGx/view. A three-minute cut: https://drive.google.com/file/d/1Loqum6Kbk1noQazJDKrrVEnHzoU3biMD/view

EvidenceSource
Across 655 large employers in 2025 the median share of staff who told their employer about a disability was 3.5%, and it did not rise. By 2026, 98% of them report accommodation programmes.Disability:IN 2025 Disability Index and 2026 results
In 2025, 22.8% of Americans with a disability were employed against 65.2% without, and 30% of those working were part time against 17%.US BLS, People with a Disability 2025, released 3 March 2026
61 of every 100 accommodations cost the employer nothing. Median one-time cost of the rest: $300.JAN employer survey, 5,406 employers, Jan 2019 to Dec 2024
More than a third of employed Canadians with a disability had an accommodation need go unmet in 2022.Statistics Canada, Canadian Survey on Disability
The GC Workplace Accessibility Passport already implements "ask once, not forever", and only for federal public servants.Government of Canada

Curb Cut is an Agentforce agent whose principal is the worker, not the employer.


The demo deployment, September 2026

The hackathon deployment below is kept as a record of what judges tested. The demo org and its number retire with the hackathon; your install has its own addresses on every row.

SurfaceAddressState
Webhttps://curbcut-dev-ed.develop.my.salesforce-sites.com/curbcutlive
Ask (the whole flow, no phone)/curbcut/asklive
Privacy/curbcut/privacylive
Terms/curbcut/termslive
SMSa Twilio number the deployment ownslive, through a relay we run beside the org
Voicethe same Twilio numberbuilt, speech in and out
Emailinbound service CurbCutInbound, address in Setup → Email Serviceslive
Slacknode channels/slack-app.mjs, Socket Mode, DM only, never sendslive in a real workspace; manifest in channels/slack-manifest.json
AgentCurb_Cut v7, Curb_Cut_Desk v5both active
Console (internal)/lightning/app/Curb_Cut_Console, home tab /lightning/n/Curb_Cut_Home, org login neededlive
Why now/curbcut/whylive
Messaging programme/curbcut/messaginglive
MCP servernode channels/mcp-server.mjs5 tools, none of which can send
Any relay or assistantPOST /services/apexrest/curbcut/v1/message/one door every channel above shares

The console is the half of this that nobody demos. Every promise the assistant makes about a person reaching a human being is worthless unless a human being has somewhere to stand. Human_Handoff__c was creating records that said "a person is picking this up" with no surface for that person.

Console surfaceWhat it is for
On Duty (app home)Three queues in the order they should be worked, with standing orders above them
Waiting for a personSomeone asked for a human and is currently getting silence
Awaiting a decisionApproved requests an employer has not answered, nearest deadline first
Needs an interpreterSigned video, waiting on a person, never machine translated
Curb Cut Overview (dashboard)Where the system is failing people, in five panels
Record feedsTwo people often work one request, so the conversation lives on the record

Nothing in the console shows a diagnosis, because there is no field for one. An operator who is asked what a person has can only answer that nobody knows and nobody can find out.

The four principles

  1. The agent's principal is the worker. Nothing leaves that the person did not personally approve.
  2. No channel is the real one. Text, voice, email and web are equal front doors.
  3. Ask before you disclose. A real answer without identifying yourself or naming anything about your body.
  4. Ask once, not forever. Standing preferences travel ahead of you.

What is deliberately not built

  • No automated sign language. ASL is a complete language; recognition is unsolved. Signed video routes to a human, with immediate text acknowledgement so nobody waits in silence.
  • No diagnosis field. It isn't encrypted or restricted. It's absent. CI fails the build if anyone adds one.
  • No phone fallback, ever. Reachable_By__c accepts a message channel or email, and the code refuses a telephone number.
  • No inference. Nothing is profiled or predicted. Every signal is one the person set.

Verification

Nothing below is asserted from memory. Each number is the output of the command beside it, and tests/sync_counts.py rewrites the published counts from the invariant suite's own output so they cannot drift.

147 Apex tests                 sf apex run test -o curbcut -l RunLocalTests554 structural invariants      python3 tests/invariants.py            ~1s, no org517 accessibility checks       python3 tests/a11y_audit.py            against the live pages122 Sa11y checks               npm run test:a11y                      Salesforce's own axe-core matcher 34 accessibility-tree checks  node tests/ax_tree_audit.mjs           what a screen reader is handed 34 controls by keyboard       node tests/keyboard_walk.mjs           real Tab presses; set CURB_CUT_SITE 21 responsible-AI checks      python3 tests/rai_self_check.py        against Salesforce's five guidelines 28 contrast checks            python3 tests/contrast_audit.py        both themes, from the tokens 16 reading-level checks       python3 tests/reading_level_audit.py 76 link and copy checks       python3 tests/link_and_copy_audit.py23/23 adversarial assertions   node tests/headless_agent_api.mjs && python3 tests/score_adversarial.py

Two of those suites break their own checks on purpose and assert each one goes red (--selftest). A suite that cannot fail proves nothing about the code it passes. The adversarial suite once scored 21 of 23, and both failures were published with their transcripts rather than averaged away; the agent's refusal wording and the scorer have since converged, and the suite passes whole. The habit stands: a miss gets reported, never smoothed.

Repository map

PathWhat
force-app/main/default/objects/9 objects, 61 fields, 31 list views, and no field for a diagnosis
force-app/main/default/classes/19 Apex classes + 15 test classes
force-app/main/default/lwc/5 Lightning Web Components, each with a Sa11y suite beside it
force-app/main/default/aiAuthoringBundles/two Agent Scripts: Curb_Cut public, Curb_Cut_Desk internal
force-app/main/default/pages/ components/ sites/the public site, six pages, anonymous
force-app/main/default/emailservices/inbound email channel
force-app/main/default/applications/ tabs/ flexipages/the internal console
force-app/main/default/reports/ dashboards/five reports and the overview dashboard, no names anywhere
channels/SMS + voice relay, Slack app (Socket Mode), MCP server, Twilio configuration
tests/the twelve suites above, plus screen_reader_walk.mjs, which writes docs/SCREEN-READER-WALK.md
deck/the deck as source; the built copy lives in docs/hackathon/
docs/hackathon/the September 2026 hackathon record: judge guides, reflections, deck, stills
legal/privacy and terms source
docs/architecture, BRD, evidence, audience, decisions, the two audit reports

Start with docs/ARCHITECTURE.md, then CRITIQUE.md, which is where the honest list of what is still weak lives.

Documentation

DocumentPurpose
docs/ARCHITECTURE.mdHow it fits together, with diagrams
docs/BRD.mdBusiness requirements and traceability
docs/ROADMAP.mdWhat is next, and what was rejected
docs/DECISIONS.mdEvery significant call and why
docs/EVIDENCE.mdEvery public figure, with source, denominator and caveat
docs/AGENT-INTERFACES.mdHeadless, MCP, agent-to-agent, and the profile we will not build
CRITIQUE.mdWhat is still wrong
FINDINGS.mdAdversarial run results
CI.mdPipeline and invariants
DEPLOY.mdDeployment

Built by Parth Sevak. Welland, Ontario, Canada. Agentforce for Good, Dreamforce 2026 · Builder Track · Abilityforce.

License

Apache 2.0. The accommodation library ships in the repo with a source citation on every row. Use it, fork it, run it in your own org.

Source: README.md at commit 1b65585

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v1.0.0LatestSep 29, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts