MCP Integration Is Live: Secure Bearer Auth for Agentic Workflows

Today, Linked.bd MCP integration is live. You can now connect MCP-capable agents with scoped bearer tokens, enforce ability-level access, and run link operations without giving broad account access.

If you want to test this immediately, create your first MCP token and connect your client in under 10 minutes.

Why this matters financially

MCP integration reduces time spent on repetitive link operations and lowers the cost of operator mistakes.

The operational value typically comes from:

Faster execution for repetitive link tasks
Less context switching between dashboard and scripts
Tighter permission boundaries for automation
Cleaner auditability via scoped token usage

When teams move link operations from manual steps to governed agent workflows, they improve both speed and control.

Quick impact model

Use this lightweight model to estimate value:

Annual impact = (hours saved x blended hourly cost) + (incidents avoided x average incident cost)

Example:

25 MCP-eligible tasks/week
6 minutes saved per task
$60 blended hourly cost
8 avoided misconfiguration incidents/year at $200 each

Calculation:

Time saved/year: 25 x 6 x 52 = 7,800 minutes = 130 hours
Labor value: 130 x 60 = $7,800
Incident value: 8 x 200 = $1,600
Total: $9,400/year

Implementation playbook

Step 1: Log in / Register

Step 2: Browse to the "API Tokens" page

Simply click this link or, navigate in this order:

Profile icon
Settings
API tokens

Step 3: Create token with MCP + Link abilities

Create a new token and enable:

Mcp Use (mcp:use)
Link abilities (links:read, and links:write if write operations are needed)

Copy the plain token value shown once. This is your bearer token. Store this token securely, it will not be shown again.

Security note: Treat MCP tokens like API keys. Never commit them to source control. Never share with someone you don't trust.

Product execution: add Linked.bd MCP to clients

Use your MCP endpoint plus bearer token:

MCP URL: https://linked.bd/mcp
Header: Authorization: Bearer <YOUR_TOKEN>

Codex

Add an MCP server in Codex with:

Transport: HTTP/streaming MCP
URL: https://linked.bd/mcp
Header: Authorization: Bearer <YOUR_TOKEN>

If your Codex setup uses JSON config for MCP servers, the shape is typically:

{
  "mcpServers": {
    "linked": {
      "url": "https://linked.bd/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_TOKEN>"
      }
    }
  }
}

Claude Code

Create an MCP server entry pointing to the same URL and include the same bearer header.

Recommended values:

Name: linked
URL: https://linked.bd/mcp
Authorization header: Bearer <YOUR_TOKEN>

Other MCP-capable tools

Any MCP client that supports:

Remote HTTP MCP endpoint
Custom request headers

can connect using the same two values (/mcp URL + bearer token).

After connection, call capabilities_get first to confirm effective tool access before write operations.

If you want to launch fast on production, create your account at linked.bd, issue a scoped token, and connect your MCP client in one pass.

Agent-assisted signup/login automation (from `/llms.txt`)

Linked.bd also supports agent-driven auth flow for MCP and API usage.

An agent can complete the onboarding flow on behalf of a user by following llms.txt (a machine-readable instruction file used by agents).

This enables guided onboarding where the agent can handle setup steps end to end, including sign-up/login support and first automation tasks, while still respecting scoped auth and account permissions.

Final takeaway

MCP integration is now a first-class path on Linked.bd.

You can start locally in minutes with a bearer token, connect Codex or Claude Code, and move to agent-assisted onboarding using the /llms.txt flow when ready.

FAQ

Do I need both `mcp:use` and `links:*` abilities?

Yes. mcp:use unlocks MCP access, while link abilities control what link tools can be called.

Can a token bypass account permissions?

No. Token abilities and account permissions are both enforced.

Can agents do login on behalf of users?

Yes, via the challenge flow in /llms.txt (/api/v1/agents/start and /api/v1/agents/verify).

What should every client call first after connecting?

capabilities_get, then read tools, then write tools.