ConnectSecure MCP

ConnectSecure + AI - for Claude, the ChatGPT desktop app, GitHub Copilot, Gemini CLI, and any agent that speaks MCP

Unofficial. Community-built Claude Code Skill and MCP server for the ConnectSecure API. Not affiliated with, endorsed by, or sponsored by ConnectSecure.

[ConnectSecure demo - animated preview]

_{▶ Watch the 30-second demo - demo data is simulated; every command shown exists in the real CLI.}

Every ConnectSecure API operation, plus a local store that answers vulnerability x asset x client questions across your whole client base in one command. Works with the AI you already use - ChatGPT (desktop app), Claude Desktop, Codex, Claude Code, Claude Cowork, and GitHub Copilot - plus Gemini CLI. Free, open source, runs on your laptop. Built for MSP owners. No code required.

Works with your agent

Security note. Local agents use the default stdio transport and need no network listener. HTTP mode (--transport http, for remote agents) refuses to start until CONNECTSECURE_MCP_HTTP_TOKEN is set to a token of at least 32 characters, rejects any request without that bearer token, serves http://127.0.0.1:7777/mcp by default, and needs --tls-cert and --tls-key for any other address. It refuses a request body over 4 MiB or one that takes longer than 30 seconds to arrive. Whoever holds the token can call every tool with your ConnectSecure credentials, including connectsecure_execute, which changes your tenant once called with confirm: true, so treat it like a password, and set CONNECTSECURE_MCP_READ_ONLY=1 on a remote host that only needs to read. See mcp-install.md.

The six agents MSP owners actually use (self-serve, works today):

The ChatGPT desktop app runs connectsecure-mcp locally and shares MCP configuration with Codex (OpenAI). ChatGPT on the web needs a remote (HTTP) server, and this connector's HTTP mode is paused until it is reprinted (see the security note in mcp-install.md), so msp-skills does not support it yet. If your team works there, a hosted gateway is the better pick; check that your AI app is on the gateway's setup list.

Also for the Microsoft and Google stacks

Big install base, but an honest heads-up: Gemini CLI runs the local binary you just installed, while Microsoft 365 Copilot and the Gemini app need a remote server, which msp-skills does not support yet.

Skill-native agents (also covered): Hermes and OpenClaw read this skill's SKILL.md directly and speak MCP - see their install sections below. Also works with Cursor, Windsurf, Cline, Continue.dev, and Zed via MCP. Full per-tool wire-up: docs/which-agent.md.

Run more than one agent? Install across all 51+ supported agents in one command: npx skills add Servosity/msp-skills@latest (requires Node.js, then run the per-skill installer for the CLI/MCP binaries). See docs/which-agent.md.

Install in 60 seconds

Fastest for Claude Desktop - one-click .mcpb

Download ConnectSecure MCP (.mcpb) - then open Claude Desktop > Settings > Extensions and select the file. One click, no JSON, no shell. (Browse every ConnectSecure release on the releases page.)

Prefer the Claude Code plugin? Add the marketplace once, then install - works immediately, no directory listing required:

/plugin marketplace add Servosity/msp-skills/plugin install connectsecure@msp-skills

Path A - paste one prompt into your AI agent (recommended)

Copy this into Claude Code, Codex CLI, or Claude Cowork:

Install the ConnectSecure Skill and MCP server from Servosity/msp-skills in this agent workspace. If this workspace uses a POSIX shell (macOS, Linux, WSL, or Bash), run bash <(curl -fsSL https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/connectsecure/install.sh). If it uses Windows PowerShell, run iwr -useb https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/connectsecure/install.ps1 | iex. Then authenticate per the README and run connectsecure-cli --help to explore.

The same prompt works in any agent that can run shell.

Path B - run the installer yourself

Windows (PowerShell):

iwr -useb https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/connectsecure/install.ps1 | iex

macOS / Linux:

bash <(curl -fsSL https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/connectsecure/install.sh)

The installer drops both connectsecure-cli (the CLI) and connectsecure-mcp (the MCP server) into your user bin path. Claude Code, Codex, and Cowork discover the Skill via SKILL.md in this directory.

Verify:

connectsecure-cli --version

Upgrade to the latest version

The installer always fetches the current release - re-run it to upgrade:

macOS / Linux:

bash <(curl -fsSL https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/connectsecure/install.sh)

Windows (PowerShell):

iwr -useb https://raw.githubusercontent.com/Servosity/msp-skills/main/skills/connectsecure/install.ps1 | iex

Claude Desktop .mcpb users: download the latest .mcpb (top of this section) and re-select it in Settings > Extensions. Claude Code plugin users: /plugin update connectsecure@msp-skills.

Add to Claude Desktop, the ChatGPT desktop app, GitHub Copilot, Gemini CLI, or another MCP client

After the installer runs, see mcp-install.md and docs/which-agent.md for the per-agent wire-up - one section per agent, including the GitHub Copilot servers key and the ChatGPT desktop app. Claude Desktop's Settings > Extensions panel is the simplest path; the MCP config block (for users who prefer editing JSON) is documented in mcp-install.md.

Install for Hermes

From the Hermes CLI:

hermes skills install servosity/msp-skills/skills/connectsecure --force

Inside a Hermes chat session:

/skills install servosity/msp-skills/skills/connectsecure --force

Hermes speaks MCP natively, so it can also use the connectsecure-mcp server directly - same install path, same env vars.

Install for OpenClaw

Tell your OpenClaw agent (copy this):

Install the connectsecure skill from https://github.com/servosity/msp-skills/tree/main/skills/connectsecure. The skill defines how its required CLI (connectsecure-cli) can be installed via the openclaw: frontmatter block.

OpenClaw isn't generally available yet; the frontmatter wiring is pre-shipped and will activate the moment OpenClaw launches.

Authenticate

ConnectSecure hosts each tenant on its own pod, so the CLI needs your pod URL plus the API client credentials from the portal (Settings > Users > (user) > API Key for the client ID and secret; Profile > API Documentation for the pod URL):

export CONNECTSECURE_BASE_URL=https://pod<N>.myconnectsecure.comexport CONNECTSECURE_TENANT=<the tenant name you type at portal login>export CONNECTSECURE_CLIENT_ID=<client id>export CONNECTSECURE_CLIENT_SECRET=<client secret>connectsecure-cli doctor

On first use the CLI exchanges those for a short-lived token (POST /w/authorize) and reuses it for at most 30 minutes. If you already hold a token, set CONNECTSECURE_ACCESS_TOKEN and CONNECTSECURE_USER_ID instead of the tenant, client ID and secret.

doctor checks config, paths, and API reachability, and with the tenant, client ID and secret set it mints a token, so a wrong secret shows up as FAIL API: client init error: ConnectSecure rejected the credentials (HTTP 401). It exits 0 even then, so scripts should add --fail-on error. A read command such as connectsecure-cli companies list --limit 1 is the end-to-end check.

Sync before you ask

exposure, portfolio, blind-spots, suppression-audit and ad-hygiene read a local mirror. Populate it once:

connectsecure-cli sync

A bare sync reads a curated set of 15 resources (companies, assets, agents, vulnerability, remediation, suppression, risk score and Active Directory reports) with GET requests only, and writes them to a SQLite file under your own user account. Credential records, integration secrets and integration-rule params, scheduled jobs (whose rows carry a password field), attack-surface results (whose rows carry creds and hashes fields), custom domains (a precaution: their certificate details are undocumented) and the agent uninstall secret are never written to it; naming one in --resources is refused before any request and stores nothing (sync --resources credentials on its own exits non-zero). Refresh with connectsecure-cli sync --full: a bare re-run only adds and updates rows, so a finding remediated since the last sync stays in the mirror (and in exposure and portfolio) until a --full run prunes it. ConnectSecure's API has no incremental filter, so --full reads the same pages a bare run does. The commands warn when the mirror is older than --max-age (30 minutes by default).

What this skill does

Full command reference: guide.md. For the AI-agent operating contract (--agent, --dry-run, when to confirm before mutating), see AGENTS.md.

What makes this different

A stateless wrapper over the ConnectSecure API answers one endpoint at a time, for one client, right now. The questions an MSP actually asks are cross-client: rank every open finding across all 40 clients at asset level, show which of them a new CVE hits, list every asset whose scan you cannot trust. ConnectSecure's per-client vulnerability report names the client but lists affected assets only by ID, so getting from a finding to the machine, its IP and its scan status is a join you would otherwise do by hand.

This skill syncs a curated set of ConnectSecure reports into a local SQLite mirror and does the joining for you. exposure joins vulnerability findings to assets and companies, portfolio merges assets, findings, suppressions, findings open past 30 days and risk score into one row per client, and blind-spots crosses discovery times, scan authentication and agent pings - instant, offline, and the AI sees the answer, not the raw data.

Two commands deliberately stay live: cve-impact (the CVE report plus the assets-by-problem report, grouped by client) and qbr (one client's risk pack). Both list what they could not fetch or attribute instead of dropping it silently. Like every read, their GET responses are cached on disk for 5 minutes, so a re-run inside that window returns the earlier answer; add --no-cache right after a fix or suppression. For scripts: with --agent, the analysis commands, list / get, sync and doctor wrap their output in {"meta": ..., "results": ...}, so parse .results; with --json, list / get use the same envelope while qbr and cve-impact print their report object at the top level. export and tail stream bare JSON records in either mode.

The pain this closes

"Vulnerability scans rarely come back clean, so MSPs are often staring at thousands of findings across dozens of customers" (ChannelPro, "Too many vulnerabilities, not enough time", updated 2026). In the same piece ConnectSecure's CEO describes the goal as narrowing 10,000 vulnerabilities to the 100 worth working in the next 30 days, which ConnectSecure approaches with AI-ready reporting in its portal. This skill works on the same problem from the AI agent you already use:

• Ranking findings across every client at asset level means matching each report's asset IDs to real machines. connectsecure-cli exposure does the join and sorts by EPSS, then CVSS, then known-exploited status.
• "Which of our clients does this CVE hit?" takes two reports and a manual grouping by client. connectsecure-cli cve-impact answers it in one command and names the fixes.
• A clean result on an asset that was not discovered in weeks, was scanned without credentials, or whose agent went quiet is unknown, not clean. connectsecure-cli blind-spots lists those assets with a reason code each.
• A suppression that expired but was never lifted, has no end date, or was approved by the person who requested it keeps a real finding out of every report. connectsecure-cli suppression-audit flags them, riskiest first.

See pain-point.md for the longer narrative.

Frequently asked questions

Does this work with ChatGPT?

Yes, in the ChatGPT desktop app. It runs MCP servers locally (STDIO), the way connectsecure-mcp runs, and shares MCP configuration with Codex (OpenAI): add connectsecure-mcp under Settings > MCP servers as a STDIO server, or add it once to Codex's ~/.codex/config.toml. ChatGPT on the web is not supported by msp-skills yet: it needs a remote (HTTP) server, and this connector's HTTP mode is paused until it is reprinted; the security note in mcp-install.md explains why. If your team works there, a hosted gateway is the better pick; check that your AI app is on the gateway's setup list.

Does this work with Codex, Cursor, Windsurf, Cline, Copilot, or Gemini?

Yes - all of them speak MCP. Cross-tool install commands are in the matrix above and the deep-dive in docs/which-agent.md.

Do I need to know how to code?

No. The recommended install is to paste one sentence into Claude Code or Codex - your agent reads SKILL.md and does the install. The fallback is a one-line installer per OS (bash or PowerShell). Neither path requires writing code. You'll enter your ConnectSecure credentials once.

Is my ConnectSecure data safe?

Your data stays on your machine. The CLI and MCP server are local binaries. The SQLite mirror and a cache of API responses (each reused for 5 minutes, kept until overwritten) sit in directories under your user account. The AI agent only sees what the CLI returns - typically a query result, not raw bulk data. Credentials are read from your environment or your agent's config; never bundled into this repo or transmitted anywhere by MSP Skills.

Where do I find my pod URL and API credentials?

In the ConnectSecure portal: Settings > Users > (user) > API Key gives you the client ID and client secret, and Profile > API Documentation shows your pod URL (https://pod<N>.myconnectsecure.com). The tenant name is the one you type on the portal login screen. Each tenant lives on its own pod, so there is no working default URL - doctor tells you when CONNECTSECURE_BASE_URL is missing.

Will this hit my ConnectSecure API rate limits?

Lightly. sync reads a curated set of reports once and stores them locally, and exposure, portfolio, blind-spots, suppression-audit and ad-hygiene then answer from that mirror without calling the API. qbr and cve-impact call the live API, with GET responses cached on disk for 5 minutes (--no-cache forces a fresh call); cve-impact reads at most --max-pages pages of 100 affected assets. By default the CLI paces itself to the server's rate-limit headers (--rate-limit).

Does it store my scan credentials?

No. Credential records (the domain and SSH accounts your scans use), integration secrets and integration-rule params, scheduled jobs (whose rows carry a password field), attack-surface results (whose rows carry creds and hashes fields), custom domains (a precaution: their certificate details are undocumented) and the agent uninstall secret are never written to the local mirror or the HTTP response cache - not by a bare sync, not by workflow archive, and not by sync --resources credentials, which stores nothing and exits non-zero by design. They are returned only to the command or agent that asked for them, which is why those reads sit in the credential tier below. The deliberate exception is export of those resources (export credentials, export integration_credentials, export integration_rules, export scheduler, export attack_surface_results, export custom_domains), which prints exactly what you asked for to stdout, or to the file you name with -o, and tail on them, which prints each poll to your terminal. To create or update credentials, integration credentials, integration rules or scheduled jobs, pass the JSON body on stdin with --stdin (wrapped as {"data": {...}}, the shape --dry-run shows for the flag form; --stdin sends it exactly as given) rather than --data-params / --data-integration-params / --data-password, so the secret stays out of process arguments and shell history; custom-domains create takes the private key as a file path (--ssl-privkey <file>). The CLI's own API token is cached under your user cache directory (owner-only permissions) for at most 30 minutes.

Will this replace the ConnectSecure portal?

No. Scanning, remediation, agent deployment and branded client reports stay in the portal. This skill answers the cross-client questions the portal makes you click through - exposure ranking, CVE blast radius, blind spots, suppression hygiene, AD hygiene - from your AI.

What does it cost?

Free. Apache-2.0 licensed. You pay only for whichever AI agent you use (Claude, ChatGPT, Codex, etc.), and that's billed by your AI provider, not by us.

Safety model

For agents that use the MCP server, set CONNECTSECURE_MCP_READ_ONLY=1: the server then refuses every non-GET call and drops sync, export, import and tail. It still answers GET calls, including the credential reads, so it is not a substitute for scoping the API user. Agents with shell access can also add --deliver webhook:<url> to any CLI command, which posts its output to that URL; the MCP server does not expose --deliver to tool calls. MCP tools also never take a local file path for the CLI to read (teach / teach-playbook notes and playbook files, import --input, the custom-domains create certificate and key, report-builder uploads); run those from the CLI.

The strongest control is the scope you grant the ConnectSecure credentials - the CLI can only do what the credentials are permitted to do. Full details, including how to lock it down, are in governance.md.

Status

Beta. Validated against the ConnectSecure API surface and being validated with MSPs running it live against their own production tenant in our weekly Build Sessions. RSVP at compoundingteams.com/build-sessions.

Standards. Conforms to the open Agent Skills spec (Anthropic, Dec 2025; 40+ agents). MCP-compatible - works with any MCP-capable agent including Hermes. OpenClaw-ready (frontmatter pre-wired, awaiting OpenClaw launch).

Maintained by Servosity. Apache-2.0 licensed. Built with CLI Printing Press. Last updated: 2026-10-13.