Remote.com
by Remote.com
Query global employment, payroll, time off and org structure from your AI assistant. 51 tools, OAuth sign-in, and the first connector in this catalogue whose OAuth scopes are split by persona — admin, employer and employee.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Remote.com via MCP
https://mcp.remote.com/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Remote.com Tools & Capabilities (51)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- This connector writes. Remote documents write tools for submitting and approving time off, creating and updating expenses, updating personal details, home address and emergency contact, managing performance reviews, configuring workflows and creating data exports. Anthropic's 51 names contain none of them, so the directory listing understates the surface.
- It does not move money. Remote states the connector cannot move money, or start payments, payouts or payroll runs, and cannot change salary or contract terms. This is a vendor statement, and it is the load-bearing difference from a payroll connector that can submit a run.
mcp:remote_admin:liteis undocumented. Remote publishes no definition of what theliteadmin scope restricts, at any of the four sources we searched. If a consent screen requests it, its reach is undetermined.- The role gate is narrower than the docs suggest. Remote's MCP documentation describes manager workflows and its example prompts include a manager section, but Remote's role matrix — updated 2026-08-12 — marks Remote MCP access as available to company owners and super admins only, and unavailable to people managers, hiring managers, onboarding managers and company viewers. Those two Remote sources disagree. Verify against your own account.
- It reads bank accounts, home addresses and emergency contacts of employees and contractors.
get_default_bank_account,list_bank_accounts,get_home_addressandget_emergency_contactare in the published list. - The transcript is the exposure Remote flags, not the server. Remote warns that your questions and its answers may be retained in the AI client's conversation history depending on that client's settings, and states it has no control over how the client processes or stores what it receives.
- Tool parameters and safety annotations were not readable. The endpoint returned HTTP 401 to an anonymous
initializeon 2026-08-18 and HTTP 406 to aGET, so names came from the directory snapshot and meanings from Remote's documentation. No tool was called. - The published lists do not agree with each other. The directory names 51 tools; Remote's own plugin skills name 42 tool-shaped identifiers with only 9 in common; and one of those skills contradicts itself about whether two cancellation tools are exposed at all. Expect the live surface to differ from every published list.
- It is not a backend integration. Remote states the MCP is not an API you call yourself, that every request runs under your interactive sign-in, and that server-to-server automation should use the REST API instead. Its General Ledger report flow is explicitly interactive — an admin drives it — with an unattended API path described as "coming soon".
- No rate limits are published for the MCP server. Remote publishes a rate limit policy for its REST API; we found none stated for MCP. Remote's own reporting skill notes large tool responses can be truncated, and recommends field projection and paging.
Frequently asked questions
Is Remote MCP secure?
Remote states the connector applies the same controls that already protect the account. Sign-in is OAuth in your own browser, so the password never reaches the AI client; every request runs through Remote's roles, permissions and Row-Level Security; traffic is encrypted with TLS. Remote notes its platform is ISO 27001 certified and that data reaching your client falls under that client's policy.
How do roles and permissions work in Remote MCP?
Remote reuses the roles you already have. Three layers apply: your Remote role, role-based permissions that decide which tools are exposed at all, and Row-Level Security that limits what each tool returns. Remote's summary is that a manager sees only direct reports and an employee sees only themselves. Nothing is granted that signing in directly would not grant.
Can I use Remote MCP? Checking access
You need an active Remote account and an MCP-compatible client. Remote tests Claude Desktop, claude.ai, Claude Code and Cursor. Nothing is installed on the Remote side. Remote's role matrix, however, marks Remote MCP access as available to company owners and super admins only, and unavailable to people managers, hiring managers, onboarding managers and company viewers.
What can my AI see when I connect it to Remote?
The same data you see when you sign in to Remote, and nothing beyond it. Remote adds a warning worth heeding: anything typed into the AI client, including parts of your questions and Remote's answers, may be stored in that client's conversation history depending on its settings. Remote advises against sharing sensitive information you do not want retained outside Remote.
What can my manager and employer see through Remote MCP?
Only what they could already see by signing in to Remote directly. Remote states a manager can ask about their direct reports' employment details and time off, plus org structure and reporting lines, where Remote already permits it. Connecting an AI client does not change the permissions that protect your record from other people.
Which AI clients work with Remote MCP today?
Remote tests Claude Desktop, claude.ai, Claude Code and Cursor. Its documentation adds that any other client should work provided it supports OAuth 2.0 with Dynamic Client Registration over streamable HTTP. For coding tools, Remote publishes a remote-for-ai plugin that configures the connection for Claude Code, Cursor, Codex and Gemini CLI.
How do I disconnect Remote MCP from my AI client?
Remove the connection in your AI client's own settings. Remote states users can disconnect at any time, and that access is revoked automatically when a Remote account is deprovisioned — so offboarding someone in Remote also cuts off whatever AI client they had connected. There is no separate token to revoke inside Remote.
Sources
- Remote MCP introduction, for authentication, prerequisites, the read/write tool categories, the security model, compatible clients and the "what it isn't" statements (HTTP 200, retrieved 2026-08-18; Markdown variant at the same path with
.mdappended,updatedAt2026-07-15).developer.remote.com/robots.txtallows this path, names its ownllms.txt, and carries noContent-Signalline. · retrieved 2026-08-18 - Remote MCP Quick Start, for setup steps, the Claude Code command, the sandbox endpoint, the multiple-account flow and the write-action test prompts (retrieved 2026-08-18,
updatedAt2026-07-17) · retrieved 2026-08-18 - Remote Agent Skills, for the
remote-for-aiplugin and its operate-vs-build split (retrieved 2026-08-18) · retrieved 2026-08-18 - Pull GL & payroll reports, for the admin-driven General Ledger flow and the not-yet-shipped API (retrieved 2026-08-18) · retrieved 2026-08-18
- Remote's documentation index, used to enumerate pages without guessing sub-URLs and to establish that
liteappears nowhere on the developer site (475 lines, retrieved 2026-08-18) · retrieved 2026-08-18 remoteoss/remote-for-aiplugin skills, for the named write tools, the confirm-before-writing rule, the PII handling rules and the self-contradiction about cancellation tools (retrieved 2026-08-18) · retrieved 2026-08-18- Remote Help Center, "What Remote MCP can and cannot do today", for the write-tool categories and the explicit cannot-move-money statement (vendor
updated_at2026-07-01, retrieved 2026-08-18) · retrieved 2026-07-01 - Remote Help Center, "How do roles and permissions work in Remote MCP?", for the three enforcement layers (vendor
updated_at2026-06-17) · retrieved 2026-06-17 - Remote Help Center, "Is Remote MCP secure?", for OAuth, TLS, ISO 27001 and revocation (vendor
updated_at2026-06-17) · retrieved 2026-06-17 - Remote Help Center, "What can my AI see when I connect it to Remote?", for the conversation-history warning (vendor
updated_at2026-06-17) · retrieved 2026-06-17 - Remote Help Center, "Can I use Remote MCP? Checking access" and "What can my manager and employer see through Remote MCP?" — and (retrieved 2026-08-18) · retrieved 2026-08-18
- Remote Help Center, "Which AI clients work with Remote MCP today?" and "How do I disconnect Remote MCP from my AI client?" — and (retrieved 2026-08-18) · retrieved 2026-08-18
- Remote Help Center, "Example prompts to try with Remote MCP" and "Six admin workflows to try with Remote MCP", for the use cases and the counsel-review caveat — and (retrieved 2026-08-18) · retrieved 2026-08-18
- Remote Help Center, "What can the different company users do on Remote?", for the role matrix whose first row is Remote MCP access (vendor
updated_at2026-08-12). Articles were read through Zendesk's public Help Center API;support.remote.com/robots.txtpermits/hc/*/articles/and carries noContent-Signalline. · retrieved 2026-08-12 - Live auth-posture check against
https://mcp.remote.com/mcp—initializereturned HTTP 401 withWWW-Authenticate: Bearer resource_metadata="https://mcp.remote.com/.well-known/oauth-protected-resource/mcp"(2026-08-18). AGETon the same host returned HTTP 406, so no public tool catalogue is served. No tool was called. · retrieved 2026-08-18 - RFC 9728 protected-resource descriptor (HTTP 200, 2026-08-18). Names
authorization_servers: ["https://api.employ.remote.com"]and the seven persona scopes. The sandbox host publishes the identical seven. · retrieved 2026-08-18 - RFC 8414 authorization server metadata (HTTP 200, 2026-08-18). Names PKCE
S256only,authorization_codeandrefresh_tokengrants, and aregistration_endpoint. It publishes noscopes_supportedfield, so the descriptor is the only scope source. The same document is served frommcp.remote.com. · retrieved 2026-08-18 - Anthropic Connectors Directory entry, for the 51 tool names, tier, endpoint and categories (snapshot dated 2026-08-16) · retrieved 2026-08-16
- Remote privacy policy — · Support
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Productivity
- Developer
- Remote.com
- Tools
- 51
- Domain
- mcp.remote.com
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.