Paste a CRM record, call notes and an email thread into your assistant. In one to three minutes, Management OS hands back a deal read: the forecast call, the risks most likely to lose or slip the deal, next moves with owners and dates, and the questions to ask the buyer. It also flags where your CRM and the buyer's own words disagree.
Every scored point quotes your own evidence, word for word.
- Sellers: walk into the pipeline review with the evidence behind every call. Know what to ask next and what could kill the deal.
- Revenue leaders: inspect any deal in one to three minutes and see where the CRM and the evidence disagree. Describe a rep's gap and get the likely cause, the interventions in order, what to say in your next 1:1, and a check on your own coaching bias.
You pay for finished work, never tokens. Plans start at $149 a month with a 30-day free trial that includes 100 credits: mgmt-os.ai/pricing.
This repo packages the Management OS connector and two workflow skills, deal-review and rep-coaching, for Claude, Claude Code, Gemini CLI, Hermes Agent, OpenClaw and other MCP clients.
Quick install
Claude Code
/plugin marketplace add Management-OS-Inc/management-os-plugin
/plugin install management-os@management-os
Claude (chat, Cowork and the desktop app): Customize > Plugins > Add marketplace > Management-OS-Inc/management-os-plugin, then install management-os.
Any other MCP client: add this server URL, then sign in with your Management OS account.
https://app.mgmt-os.ai/api/mcp
Then ask for a read and paste your evidence:
Use Management OS to score this deal. Here's the opportunity record, my notes from Tuesday's call and the last email thread.
Requires: a Management OS account with agent use: an Agent plan, or an app package with Agent Access. Work is paid in credits; see pricing.
New here? Start with the sample deal. Steps for each platform are below, and the full setup guide is at mgmt-os.ai/help/ai-assistants.
Try it on a sample deal
Everything in this deal is made up: the company, the people and the numbers. Once you're connected, paste it into your assistant. A deal read takes one to three minutes and uses 5 credits.
Copy the sample deal
Use Management OS to score this deal.
Deal: Tallowmere Logistics - Dispatch Platform
Account: Tallowmere Logistics
Amount: $184,000
Stage: Negotiation
Close date: 2026-10-31
Evidence 1. Salesforce opportunity record, updated Oct 10
Stage: Negotiation. Forecast category: Commit. Amount: $184,000 ARR (3-year term). Close date: 2026-10-31. Owner: Sam Ortega. Champion: Dana Whitfield, VP Operations. Decision maker: Marcus Lee, CFO. Competitor: Brindlepath; current process runs on spreadsheets. Next step: Send final order form. CFO signature expected this week. Last activity: Oct 6 pricing call.
Evidence 2. Gong transcript, pricing call, Oct 6
Sam Ortega: Thanks for making time, Dana. You've seen the final pricing. Where do we stand?
Dana Whitfield: I'm all in on this. My dispatch team is begging for it. We lose two or three hours a day rebuilding routes by hand.
Sam Ortega: Great. Has Marcus seen the number?
Dana Whitfield: Marcus hasn't seen the final number yet. I'll walk him through it after our board meeting on the 22nd.
Sam Ortega: Can we get this signed by the 31st?
Dana Whitfield: That's the goal. I can't promise it. Security still needs to sign off. Priya's team has the questionnaire.
Sam Ortega: Anything else that could get in the way?
Dana Whitfield: Brindlepath came in about thirty percent lower. Marcus will ask about that, and I need a better answer than features.
Sam Ortega: Understood. I'll send the order form Monday.
Evidence 3. Gmail thread, "Re: Order form and security questionnaire", Oct 9 to 10
Priya Raman (IT Security), Oct 9: Sam, we received the security questionnaire. Our review queue is running about three weeks right now. We'll reach out with questions.
Dana Whitfield, Oct 10: Sam, quick heads-up. Marcus asked whether this can come out of the Q1 budget instead of this year's. I told him I'd find out what that does to pricing. Can you send options?
Evidence 4. Rep notes, Oct 10
Dana is a strong champion and wants this badly. Haven't met Marcus yet. Plan: send order form Monday, push for signature by 10/31. Feeling good, 90%.
The CRM says Commit and the rep says 90%. The buyer's own words say otherwise. Part of a real read of this deal:
Tallowmere is not a $184,000 October close. The CFO hasn't seen the number, security needs about three weeks and he's asking about Q1 budget. Take it out of commit.
CFO not engaged and leaning toward Q1. The rep hasn't met Marcus. He hasn't seen the price and won't until after the 22nd. He is already asking to move the spend to Q1 budget. Without him there is no signature, and October 31 depends on it. Evidence: "Marcus asked whether this can come out of the Q1 budget instead of this year's." (Gmail)
Next move: Get the champion to set a meeting with Marcus before the 22nd. Success means a date on the calendar and Marcus stating the budget year. Owner: AE with champion. By: before Oct 22.
The full read adds two more risks, two more moves, the questions to ask the buyer, where the sources disagree and all 15 factors.
Then ask for the coaching (3 credits):
Use Management OS to help me coach Sam on this.
Three of his five Commit deals slipped last quarter.
None of those three had a meeting with the economic buyer before Commit; both deals that closed did.
I already told him to be more conservative, and he committed this one at 90% the next week.
Primary cause: information. Sam has no hard rule for what Commit means, only a mood. His own data has the rule in it: 2 of 2 deals with an economic-buyer meeting closed, and 3 of 3 without one slipped.
Open the 1:1 with: "Before we get to Tallowmere, I want to look at your Q3 Commits with you. Five deals, two closed, three slipped."
Check your own bias: The manager has already reached for the vague fix ("be more conservative"), and forecasting training is the natural next step. Both miss the cause.
Excerpts from real runs of this sample. Wording varies from run to run.
Then try it on your own deal. Pick one you're about to forecast. Paste the CRM opportunity record, the latest call notes or transcript, the email thread with the buyer and your own notes, each as it is, and say where it came from. Don't summarize: the read quotes your evidence, so the buyer's own words matter. More sources make a sharper read. A read covers up to 10,000 words of evidence.
What Management OS adds
- Proof on every point. Each scored point rests on a word-for-word quote from your evidence. Points the scoring passes disagree on are scored conservatively, and every read is checked automatically before it reaches you.
- One standard for every deal. The same 15 factors of opportunity health for every deal and every rep, so one rep's Commit means the same as another's.
- Built for whoever owns the number. Coaching for the leader: the likely cause of a rep's gap, the interventions in order, the words for your next 1:1 and a check on your own coaching bias.
- Context from the app. If the deal is in Management OS, say so and the read adds its synced CRM record. In the app, Management OS also keeps deal pages, contact dossiers and meeting prep current from your CRM and connected tools.
What you get
| Tool | What it does |
|---|---|
score_deal | An evidence-cited read on one deal: the forecast call, the risks most likely to lose or slip it, next moves with owners and dates, questions to ask the buyer, and where your sources disagree. Every scored point quotes the evidence you shared. |
coach_rep | The likely cause of a rep's performance gap, interventions in order, how to run the next 1:1, what to say, and a check on your own coaching bias. |
explain_canon | What Management OS evaluates and how it handles your data, in plain language. |
get_usage | Your workspace's Management OS agent activity and credit balance. |
| Skill | Use it when |
|---|---|
| deal-review | You want to know how a deal is really doing, whether to commit it, or what to do next. |
| rep-coaching | A rep is behind and you need the cause and the next 1:1. |
No tool sends messages, changes CRM records or deletes data.
Install by platform
Server URL for every client: https://app.mgmt-os.ai/api/mcp
Claude: chat, Cowork, mobile and agents
- Plugin (connector + skills): Customize > Plugins > Add marketplace >
Management-OS-Inc/management-os-plugin, then install management-os. - Connector only: Settings > Connectors > Add custom connector > paste the server URL > Connect.
Claude Code
/plugin marketplace add Management-OS-Inc/management-os-plugin
/plugin install management-os@management-os
ChatGPT: in review
Management OS is in review for ChatGPT's app directory. Once it's listed, find Management OS there and select Connect.
Grok
On grok.com: Connectors > New Connector > Custom > paste the server URL > sign in.
Gemini: CLI, Gemini app and Gemini Enterprise
- Gemini CLI:
gemini extensions install https://github.com/Management-OS-Inc/management-os-plugin, then/mcp auth management-os. - Gemini app (personal Google accounts in the US): Settings > Connected apps > Custom apps > add the server URL.
- Gemini Enterprise: email support@mgmt-os.ai for an OAuth client for your organization.
Hermes Agent
hermes plugins install Management-OS-Inc/management-os-plugin
hermes plugins enable management-os
hermes mcp login management-os
OpenClaw
openclaw mcp set management-os '{"url":"https://app.mgmt-os.ai/api/mcp","transport":"streamable-http","auth":"oauth"}'
openclaw mcp login management-os
Skills in other agents
npx skills add Management-OS-Inc/management-os-plugin
Any other MCP client: Cursor, VS Code, GitHub Copilot, Codex and more
Point it at the server URL. Streamable HTTP. Sign-in is discovered automatically (OAuth 2.1 with a Client ID Metadata Document or Dynamic Client Registration).
Data and privacy
The tools receive only what your assistant sends in a call: the deal or rep details and the evidence for that request. Evidence is used for that call only. Management OS keeps a one-way fingerprint of it, and keeps each read, which quotes your evidence, for up to 24 hours before clearing it. Management OS does not train models on customer data. Full details, including our AI provider: Privacy Policy, Section 12.4.
Found a security issue? Report it privately: see our security policy.
What's in this repo
| File | For |
|---|---|
.claude-plugin/plugin.json (declares the MCP server), .claude-plugin/marketplace.json | Claude and Claude Code |
plugin.json, mcp.json | Clients that read Agent Plugins 1.0 manifests |
gemini-extension.json, GEMINI.md | Gemini CLI |
skills/ | Agent Skills, read by clients that support them |
docs/images/ | Images for this README |
These files hold no code and no credentials. The tools run on Management OS servers.
Support
support@mgmt-os.ai · mgmt-os.ai/help/ai-assistants
License
The files in this repo are MIT licensed. See LICENSE. Management OS itself is provided under its Terms of Service.