What is the Buildforce MCP endpoint?

The Buildforce MCP server is at https://mcp.buildforce.io/mcp. It uses OAuth 2.1 with PKCE; discovery metadata is published at https://mcp.buildforce.io/.well-known/oauth-authorization-server.

How do I connect Buildforce to Claude, ChatGPT, or Gemini?

In Claude: Settings → Connectors → Add custom connector → paste the endpoint → approve sign-in. In Claude Code: claude mcp add --transport http buildforce https://mcp.buildforce.io/mcp. In ChatGPT (Business/Enterprise) and Gemini Enterprise: Settings/Admin → add MCP server → paste the endpoint. In Microsoft Copilot Studio: add a Model Context Protocol tool with the same endpoint.

Is it safe to let the AI change my production org?

Yes. Read-only tools run freely; state-changing tools require explicit confirmation. Production deployments are routed through a sandbox or pipeline first, require explicit production confirmation, and ship with rollback. Per-tenant scoping is enforced at the gateway.

Buildforce MCP Connector — Documentation

What it is

The Buildforce connector exposes your Salesforce, ServiceNow, and HubSpot operations capabilities to the AI client you already use. Connect once with OAuth and your AI can triage your backlog, diagnose adoption, assess deployment safety, quantify tech debt, and deploy approved fixes — grounded in your real org.

Endpoint

https://mcp.buildforce.io/mcp/health-check

Authorization: OAuth 2.1 + PKCE. Discovery metadata at https://mcp.buildforce.io/.well-known/oauth-authorization-server.

New to Buildforce? Start here

The connector signs you in to an existing Buildforce account — it can't create the account, confirm your email, or connect a Salesforce org for you, because each of those steps happens in your browser. If your AI tells you no account or no connected org was found, do this once, then return to your AI:

Create a free account: buildforce.io/auth — email + password, no credit card. Your workspace is created automatically.
Connect a Salesforce org: open buildforce.io/platforms and click Connect. The free tier includes up to 2 connected platforms and free health checks.
Back in your AI: sign in through the connector with the same email and ask "What's broken in my org?"

Already have an account? Skip to connecting your AI client below. Inside your AI you can ask the get_started tool any time to see exactly which of these steps you still need.

Connect in 60 seconds

Claude / Cowork: Settings → Connectors → Add custom connector → paste the endpoint → approve sign-in.
Claude Code: claude mcp add --transport http buildforce https://mcp.buildforce.io/mcp/health-check
ChatGPT (Business/Enterprise): Settings → Connectors → Add MCP server → paste the endpoint (admin enables for Enterprise).
Gemini Enterprise: Admin console → add MCP server → paste the endpoint.
Microsoft Copilot: Copilot Studio → add a Model Context Protocol tool → paste the endpoint.

Available tools

Read-only (run freely):

get_health_check_status — progress and results of a scan
get_health_recommendations — prioritized fixes
calculate_cost_impact — dollar cost of tech debt
predict_health_trends — where org health is heading
get_industry_benchmarks — peer comparison
analyze_dependencies / resolve_dependencies — change blast radius and safe order
assess_release_readiness — is a change safe to ship
analyze_process_flows — automation redundancy and conflicts
analyze_compliance, get_anomaly_analysis, get_metric_history

State-changing (your AI will confirm first):

run_saas_health_check — start a full scan of a connected org
configure_anomaly_detection — set monitoring thresholds
Deployment tools — metadata deploy with rollback

The technical-consultant skill

Buildforce ships a portable consultant skill for Claude Code, ChatGPT, and Gemini that turns these tools into guided workflows: org triage, adoption diagnosis, deployment safety, data-model changes, and the cost-of-debt business case. Ask your AI: "Connect our Salesforce org and tell me what's broken."

Security

OAuth 2.1 with PKCE; tokens encrypted at rest; per-tenant scoping enforced at the gateway; production changes require explicit confirmation and ship with rollback. See our privacy policy.