Ask your Beeline data. Get answers. Take action.
Ask in plain language. Get answers from live Beeline data. Take supported actions, all within existing permissions.
Beeline MCP (Model Context Protocol) is a secure, standards-based connection between Beeline Enterprise and the AI assistants and agents an organization already uses. Instead of switching between applications, a hiring manager can ask an assistant, "Show my pending timesheet and expense approval counts," and get an answer drawn directly from Beeline. Supported actions, such as approving a timesheet or favoriting a candidate, can be completed in the same conversation. The result is faster answers and quicker decisions, with the permissions and approval workflows already trusted in Beeline staying firmly in place.
Early access begins October 31, 2026. This article covers what is available at launch, how access is governed, and what comes next.
Jump to what you want
At a glance
| Available at launch | What this enables next (directional) |
|---|---|
| Hiring manager and approver workflows across approvals, requests, candidates, assignments, SOWs, and projects | Additional hiring manager use cases, such as creating requests and managing assignments |
| Bring-your-own-agent: connect the MCP-compatible assistant the organization already uses | More personas, first-party Beeline agents, and a broader bring-your-own-agent ecosystem |
| User-scoped access: every request runs as the signed-in user, under that user's role and permissions | Agent identity and on-behalf-of transactions, so actions are attributed to both person and agent |
What is MCP, in plain terms?
Model Context Protocol is an open standard that lets AI assistants discover and use the tools an application offers. Think of it as a governed front door to Beeline: an assistant sees a defined set of tools, calls the ones it needs, and Beeline decides what each signed-in person is allowed to see and do. Because the standard is open, organizations are not locked into a single AI vendor or a Beeline-owned studio. The organization brings the agent; Beeline brings the data, the actions, and the guardrails.
Available at launch
The launch tool set is focused on the hiring manager and approver experience: finding what needs attention, reviewing it, and acting on it. The table below shows what is possible in each area. Read-only means the assistant can look but not change anything.
| Area | What you can do | Type |
|---|---|---|
| Approvals | See how many approvals are waiting, list pending timesheet and expense approvals, and view each approval's steps. Check how a timesheet's estimated cost adds up, day by day. Approve or reject a step, approve several at once, or complete an assigned task. | Read and action |
| Requests | List requests, requests delegated to the user, and requests visible through organizational relationships, with status, dates, and the actions available. Place a request on hold, reopen it, or cancel it. | Read and action |
| Candidates | Review new candidates, group them by request, and see counts by status. Read and add notes, mark candidates as seen, favorite them, and look up resume attachment details. Approve a candidate's current step (for example, offer or onboarding) or reject a candidate. | Read and action |
| Assignments | List assignments, assignments delegated to the user, and assignments visible through organizational relationships. | Read only |
| SOWs | List SOWs with owner, status, supplier, dates, and amount. Check pending amendments and supplier contacts, and generate an export document for a SOW. | Mostly read only |
| Projects | List projects with manager, status, supplier, budget, and dates. Add project notes and view or update project attachments. | Read and action |
Need ideas for what to ask?
See Beeline MCP: Example prompts and use cases for sample questions and common scenarios.
Looking for the latest tools and actions?
Available tools and actions depend on your organization's configuration and your permissions. Visit the Beeline MCP Portal for the current list available to you.
Not part of the launch tool set
To set clear expectations, the following are not included at launch: creating new requests or SOWs, amending, extending, or terminating assignments, and support for roles beyond hiring managers and approvers. These are the areas Beeline is building toward, described below.
How governance works
Beeline MCP does not create a new set of rules for AI. It connects assistants to the permission model Beeline already enforces for every user. In practice, that means five things.
| Control | How it works |
|---|---|
| Actions run as the signed-in user | Access is user-scoped. The assistant authenticates through the organization's identity provider and acts as the signed-in user, not as a shared system account. What the assistant can see is limited to what that user can see, based on Beeline role-based access and permissions. |
| Permissions and workflow state still apply | Every action is subject to the user's permissions and to the state of the item. For example, a user can approve a step only when it is waiting on that user, and a request offers only the actions its current status allows. Beeline enforces this inside the application, so an assistant cannot do anything the user could not do directly in Beeline. |
| Irreversible changes are confirmed first | Actions that cannot be undone, such as rejecting a candidate or an approval step, or cancelling a request, require explicit confirmation before they run. Reading information never changes anything. |
| The organization decides who connects | MCP is enabled for a Beeline Enterprise instance by Beeline through a Service Request. The organization chooses which approved assistants and agents connect, and users sign in with the credentials they already use. |
| People stay the decision makers | Assistants help find, summarize, and act. Decisions such as approving spend or rejecting a candidate remain with people, and every action is attributable to the person it was performed for. |
TIP: AI assistants are great at summarizing, but Beeline remains the system of record.
What this enables next
Launch is the foundation. The same governed connection is designed to grow into the areas below. These are directional themes, not commitments, and timing and scope are subject to change.
| Theme | What it will make possible |
|---|---|
| Agent identity | Every action carries both the identity of the person and the identity of the agent acting for that person, giving customer teams clearer attribution and oversight as agent use grows. |
| On-behalf-of transactions | Assistants and agents acting for another person under proper delegation, for example an approver's delegate or a program team member acting for a hiring manager, always within that person's authority. |
| More hiring manager use cases | Going beyond review and approval to originating work: creating requests, managing assignments (amend, extend, terminate), submitting SOWs, and additional approval types. |
| Additional personas | Bringing the same experience to other roles in the program, for example MSP and supplier users, each with their own role-based access. |
| First-party agents | Beeline-built agents that use the same tools and the same controls, so ready-made experiences can sit alongside agents organizations bring themselves. |
| Broader bring-your-own-agent ecosystem | More AI platforms and partner-built agents connecting to Beeline through one governed layer, without rebuilding logic in a vendor-specific studio. |
Who it's for and how to get started
| Topic | Details |
|---|---|
| Who | Beeline Enterprise customers participating in the Beeline MCP early access program, and the teams and partners connecting an approved AI assistant or agent to a program. |
| How to enable | Adding Beeline MCP to a Beeline Enterprise VMS requires a Service Request and a statement of work before final production enablement. Beeline enables and configures MCP for participating organizations. |
| Service Request details |
|
| Cost | Early adopters receive Beeline MCP trial access for their project environment at no additional cost from October 31, 2026, through December 31, 2026. Future MCP pricing and packaging are being finalized. |
| Documentation | Tool details, supported clients, and configuration instructions are available on the Beeline MCP portal. |
Frequently asked questions
Which AI assistants can be used?
Beeline MCP works with MCP-compatible assistants and agents, for example, Claude, ChatGPT, Microsoft Copilot, Gemini, enterprise agent platforms such as Moveworks®, and proprietary AI front ends that support MCP connections. Supported clients and configuration options vary by implementation, so the Beeline representative can confirm current compatibility.
Can an assistant see data that a user cannot see?
No. The assistant acts as the signed-in user and is limited to the information and actions permitted by the user's Beeline role and permissions.
Can an assistant approve or reject items?
Only within the signed-in user's permissions and the item's workflow state. Irreversible actions, such as rejecting a candidate or canceling a request, are confirmed before they run.
Does MCP replace Beeline's APIs?
No. APIs and MCP serve different purposes and work together. See Connecting to Beeline: Integration options and how MCP differs from APIs.
Is Beeline MCP available to all customers?
Not yet. It is initially available to participating Beeline Enterprise customers through the early access program.
Learn more
- Beeline MCP: Example prompts and use cases
- Connecting to Beeline: Integration options and how MCP differs from APIs
- Beeline MCP portal for tools, supported clients and setup
| Persona | Modules AI Assistants & MCP | Documentation release Beeline Enterprise | Q4 2026 | Feedback? Email us: |