Claude and ChatGPT connector · AI-first

Nonprofit comp benchmarks, right inside Claude or ChatGPT.

Connect your CauseComp account to Claude once, then ask for executive compensation benchmarks and staff salary ranges in plain English — powered by the same engine behind our web tools. 1,100,000+ executive records, 850+ workforce roles, and §4958 comparability data.

Go Professional to connect → Explore the web tools

One-click connect

The CauseComp connector works wherever you use Claude — connect it once and it follows your Claude account: claude.ai, the Claude desktop app, and Claude's integrations inside other tools where connectors are enabled. There are no API keys to copy for the one-click path — you sign in with your CauseComp account and approve access.

1

Open Connectors

In Claude, go to Settings → Connectors, click Add, then Add custom connector.

2

Fill in two fields

Name: CauseComp

Remote MCP server URL: https://mcp.causecomp.org/mcp

Leave everything under Advanced settings — OAuth Client ID and OAuth Client Secret — blank; sign-in comes next, no keys to paste. Then click Add.

3

Sign in & allow

Click Connect on the new CauseComp row, sign in with your CauseComp account, and Allow — you're benchmarking.

Using ChatGPT

Using ChatGPT instead? A ChatGPT plugin is coming soon. In the meantime, your team can connect CauseComp to ChatGPT today by turning on developer mode and adding CauseComp as a custom MCP connector. It works in ChatGPT on the web on Go, Plus, Pro, Business and Enterprise plans; on Business and Enterprise, a workspace admin sets it up for the workspace first.

The MCP server URL is the same one Claude uses: https://mcp.causecomp.org/mcp

In Claude or in ChatGPT, the connector requires a CauseComp Professional-plan account, and you sign in with OAuth when you connect.

What you can ask

Ask in plain English. Claude calls CauseComp and returns the percentiles, the data vintage, and the disclaimer — and when a state, sector, or role isn't recognised, it gets the valid list back and asks again instead of guessing.

Executive benchmark“What's the median total compensation for an Executive Director at a $5M human-services nonprofit in California?”
Workforce range“Give me the P25–P75 salary range for a Registered Nurse in the Chicago metro.”
Comparables for a board file“Pull the §4958 comparables set for that CEO benchmark, with the composition line, so I can put it in our board packet.”
Sanity-check a role“Is $145,000 reasonable for a Development Director at a mid-size arts organization?”
Use your own peers“Use our own peer list — here are twelve EINs — and show me what each of them paid their CFO, with the EIN on every row.”
Peers across sectors“Rank CEO comparables for a $25M organization across Health and Human Services, not just our own sector.”
Every officer at once“For a $10M education nonprofit in Ohio, give me the comparables set for all officer roles in one go.”
Nonprofit vs private“Nurses at nonprofits versus the private sector in Denver, senior level — and tell me which geography the number actually used.”
Narrow the cohort“Executive Director pay for 501(c)(3) youth-development organizations — subsector P20 — around $2.5M, in the Dallas metro, with the pay-component breakdown.”
The correction“Benchmark a CFO in Califoria” — Claude gets the valid state list back, asks which you meant, and never serves the national number by mistake.
The board packet“Generate the board packet for our CEO benchmark — the §4958 PDF I can put straight in the meeting file.”
Name to EIN“Find the EINs for these eight organizations, then use them as our peer set.”
Cite the method“How was that median computed? Quote the methodology section that covers it.”
Project to a date“Project that Executive Director benchmark to June 30, 2028 and tell me the rate you used.”

What's included

Professional  The connector is a Professional feature. Through it, Claude reaches the full toolset — both engines and the compliance layer:

Executive benchmarks

Officer and leadership pay from IRS Form 990 filings, by budget size, sector, state or metro, with P10–P90 and pay components (base, bonus, other, retirement and deferred, nontaxable) modelled as shares of the total from the filings that itemize them. Narrow to a 501(c) type or an NTEE subsector such as P20. Project the figures to a date you choose, at CauseComp's measured pay growth rate or one you supply — a projected figure always says that it is.

Workforce benchmarks

Broad-based staff pay across 850+ roles and ~390 metros, from BLS wage data — nonprofit or private-sector, by seniority level and headcount, with the geography actually used (metro, state, or national) stated on every answer.

§4958 comparables

The §4958 comparables set for the same inputs as an executive benchmark, with what each reported, and a composition line on every set: how many peers, how many in your state and sector, and the size of the national pool.

Your own peer list

Hand Claude your board's comparables by name or EIN, in any mix, and get what those organizations filed for the role instead of a modelled cohort. Ask for EINs on every row, and add a city to tell same-named organizations apart.

Peers across sectors, several roles at once

Rank the peer set across a list of sectors or all of them, for one role, a list of roles, or every officer role, in a single call. The benchmark itself stays the sector you asked for.

Role, sector, state & metro lookups

Vocabulary tools so Claude turns “head of HR”, “the Bay Area”, or “upstate New York” into the exact role, sector, state, or metro the engine accepts — and never guesses when it can't.

Board packet

Ask for the board-ready §4958 PDF itself, not just the numbers: the same document the website generates, with the modeled benchmark, the comparables and their own P10–P90 ladder, the composition sentence, and the disclaimer. Claude hands you the file.

Find an organization by name

Look up nonprofits in the comparables index and get each one's EIN, city, state, sector, revenue band and latest filing year. Up to 20 candidates, alphabetical and unranked, so you choose — then hand those EINs straight to a peer set.

Cite the methodology

Ask how a number was built and Claude quotes the published methodology by section, live from the same /methodology page you can read yourself. There is no second copy to drift from it.

The connector and the API are Professional features. See plans & pricing

Other MCP clients and your own code

Calling the API from your own code, or using a Model Context Protocol client other than Claude and ChatGPT? Create an API key on your account page. Inside Claude — claude.ai, the desktop app, or Claude's integrations in other tools — you don't need a key: the one-click connector is the easiest path. ChatGPT in developer mode doesn't need one either: it connects with OAuth sign-in. See the API & connector docs for setup.

Secure by design

Sign in, don't share secrets

You connect with OAuth, in Claude or in ChatGPT. Neither assistant sees your CauseComp password, and CauseComp never sees your Claude or ChatGPT account.

Revoke anytime

Disconnect the connector or revoke any API key from your account page whenever you want; access stops immediately.

Per-key daily quotas

Each key carries a daily request quota, so a shared or leaked key can't run away with your usage.

Entitlements resolve live

Every request checks your current plan — access always matches your subscription, nothing is baked into a token.

Questions

Connector FAQ

What is the CauseComp Claude connector?

A connector that lets Claude answer nonprofit compensation questions using your CauseComp account — executive pay from IRS Form 990 filings and workforce pay from BLS data. Add it once in Claude's connector settings and sign in with your CauseComp account; Claude then calls the same benchmarking engine that powers the web tools.

How do I connect CauseComp to Claude?

In Claude, open Settings → Connectors, click Add, then Add custom connector. For Name enter CauseComp, and for Remote MCP server URL enter https://mcp.causecomp.org/mcp. Leave the Advanced settings fields (OAuth Client ID and OAuth Client Secret) blank, then click Add. Click Connect on the CauseComp row, sign in with your CauseComp account, and Allow — no API keys to copy for the one-click connector.

Which plan do I need?

The Professional plan. The connector itself and the §4958 comparables data are Professional features; executive and workforce benchmarks are included from Essential up. The connector resolves your live plan on every request.

Does it work with Claude Desktop and other AI tools?

Yes. The one-click connector works in the Claude desktop app the same way as on claude.ai — connect once in Settings → Connectors and it follows your Claude account, including Claude's integrations in tools like Word and Excel where connectors are enabled. For MCP clients other than Claude and ChatGPT, create an API key on your CauseComp account page.

Is it secure? Can I revoke access?

You sign in with OAuth, so passwords are never shared. Revoke the connection or any API key at any time from your account page, and each key has a daily request quota.

Go Professional → Questions? Contact us

Claude is a trademark of Anthropic, PBC. CauseComp is not affiliated with or endorsed by Anthropic. ChatGPT is a trademark of OpenAI. CauseComp is not affiliated with or endorsed by OpenAI.