Research Project This is a free AI research project. No warranties, SLAs, or company associations. Learn more

Integration

Claude in Excel.
Your policies. Your budget. Your audit trail.

Route Claude for Excel & PowerPoint through BrainstormRouter. Every cell formula and slide generation flows through your governance layer — budget limits, PII scanning, model routing, and full observability.

The Problem

Direct Connections Skip Governance

When Excel users connect directly to Anthropic, every request bypasses your security policies. No budget limits, no audit trail, no PII scanning. One runaway spreadsheet can burn through your API budget overnight.

Direct to Anthropic

  • No per-user budget limits
  • No audit trail of requests
  • No PII scanning on inputs
  • No model routing optimization
  • No central credential management
  • No visibility into usage patterns

Via BrainstormRouter

  • Per-user and per-key budget enforcement
  • Full request/response audit log
  • PII scanning before provider relay
  • Auto-route to optimal model per task
  • Keys stored server-side, never in add-in
  • Real-time usage dashboard and alerts
How It Works

Three Steps to Governed AI in Office

1

Point the add-in at BrainstormRouter

In the Claude add-in settings, replace https://api.anthropic.com with https://api.brainstormrouter.com. Use your BR API key in the x-api-key header.

2

BrainstormRouter authenticates & applies policies

Every request hits Guardian Intelligence: budget check, PII scan, rate limit, model selection. The Anthropic Messages API facade translates between formats transparently.

3

Response streams back with full observability

SSE streaming works identically to direct Anthropic. Every response includes cost tracking, model info, and audit headers. The user experience is unchanged.

API Contract

Three Endpoints. Full Compatibility.

The Claude add-in requires exactly three endpoints. BrainstormRouter implements all three with the Anthropic Messages API format.

Method Endpoint Purpose Auth
GET /v1/messages Credential validation (add-in checks on connect) x-api-key
GET /v1/models Model list in Anthropic format ({data: [{id, ...}]}) x-api-key
POST /v1/messages Chat completions with SSE streaming x-api-key
Enterprise Value

Six Layers of Governance

Central Policy Control

One policy engine governs all Office AI usage. Model restrictions, content filters, and rate limits apply uniformly across every Excel and PowerPoint user.

Per-User Budgets

Set daily and monthly spend limits per API key. Guardian Intelligence predicts costs before execution and blocks requests that would exceed budgets.

Full Audit Trail

Every request logged with user identity, model used, tokens consumed, cost incurred, and response time. Export to your SIEM or compliance system.

Intelligent Routing

Simple cell formulas route to fast, cheap models. Complex document analysis routes to Claude Opus. Thompson sampling optimizes automatically.

PII Scanning

Every request passes through the streaming firewall. Detect and block PII, credentials, and sensitive data before it reaches the provider.

MCP Tool Bridge

Extend Excel AI with 65 MCP tools — memory queries, database lookups, web searches — all governed by the same RBAC and audit layer.

Troubleshooting

Common Issues & Fixes

Production-tested error decoder from real Claude for Excel deployments.

Symptom Cause Fix
Add-in says "Invalid API Key" GET /v1/messages must return 200 (not 405) Ensure BR API key has Messages API access enabled
No models appear in dropdown GET /v1/models must return Anthropic format, not OpenAI BR auto-detects x-api-key header and returns Anthropic format
Streaming fails mid-response SSE event format mismatch (add-in expects event: message_delta) Ensure you're hitting /v1/messages not /v1/chat/completions
CORS errors in browser-based add-in Add-in origin not in allowed CORS list BR allows * origins for /v1/messages endpoints
Budget exceeded mid-session Guardian blocks when daily limit is hit Increase key budget in dashboard or wait for reset

5-Minute Setup

1

Create a BrainstormRouter API key

Go to Dashboard → API Keys → Create Key. Set a daily budget limit appropriate for the user.

2

Add your Anthropic key to BrainstormRouter

Dashboard → Provider Keys → Add Anthropic. The key stays server-side; users never see it.

3

Configure the Claude add-in

In Excel/PowerPoint: Claude add-in settings → API Base URL: https://api.brainstormrouter.com → API Key: your BR key.

4

Test the connection

Try a simple prompt in Excel. Check the Dashboard for the request showing up with cost and model info.

5

Roll out to team

Create per-user API keys with individual budgets. Each key gets its own audit trail and usage metrics.

Get Your API Key