> ## Documentation Index > Fetch the complete documentation index at: https://docs.interchange.io/llms.txt > Use this file to discover all available pages before exploring further. # Sandbox Mode > Test the full media buying lifecycle without real spend ## Overview Sandbox mode lets you test the full media buying lifecycle — product discovery, campaign creation, creatives, execution, and delivery — without real platform calls or spending real money. Just create a sandbox advertiser and everything else is handled automatically. Validate your workflows end-to-end before going live. No real bids, no real spend. Create a sandbox advertiser and you're done. Account routing and environment isolation are handled for you. *** ## How It Works Sandbox is **account-level, not per-request**. The seller provisions a dedicated sandbox account, and every request using that account is automatically treated as sandbox. This eliminates the risk of accidentally mixing real and test traffic in a multi-step flow. When you create an advertiser with `sandbox: true`: * All discovered accounts for that advertiser are sandbox accounts * The correct sandbox account is automatically injected into every ADCP call — `create_media_buy`, `get_media_buy_delivery`, and `get_products` * Delivery and reporting data are fully scoped to the sandbox environment * Responses contain simulated but realistic data There is no additional configuration needed. You use the same endpoints and workflows as production — the only difference is the `sandbox: true` flag on the advertiser. ### Seller ad-server setup For an ad-server-backed storefront, Interchange also needs a dedicated advertiser/account inside the seller's ad server. Keep it separate from every production advertiser. The seller can either: * Create or designate a sandbox advertiser/account and assign it to the Interchange service account, then map it as the storefront's sandbox advertiser; or * Grant the Interchange service account permission to create advertisers so Interchange can provision `Interchange - Sandbox` automatically. Until one of those paths is complete, Interchange will recommend the setup in storefront readiness and may send a reviewed seller Nudge. This recommendation does not block live selling, but smoke tests will not run through a production or default advertiser as a fallback. For protocol-level details on how sandbox mode works, see the [AdCP Sandbox documentation](https://docs.adcontextprotocol.org/docs/media-buy/advanced-topics/sandbox#sandbox-mode). *** ## Creating a Sandbox Advertiser ### Via API Set `sandbox: true` in the create advertiser request body: ```bash curl theme={null} curl -X POST "https://api.interchange.io/api/v2/buyer/advertisers" \ -H "Authorization: Bearer your-api-key" \ -H "Content-Type: application/json" \ -d '{ "name": "Acme Corp (Sandbox)", "description": "Integration testing advertiser", "sandbox": true }' ``` ```typescript TypeScript theme={null} const response = await fetch('https://api.interchange.io/api/v2/buyer/advertisers', { method: 'POST', headers: { 'Authorization': `Bearer ${process.env.SCOPE3_API_KEY}`, 'Content-Type': 'application/json', }, body: JSON.stringify({ name: 'Acme Corp (Sandbox)', description: 'Integration testing advertiser', sandbox: true, }), }) const { data } = await response.json() console.log('Sandbox advertiser:', data.id, data.sandbox) // sandbox: true ``` ```python Python theme={null} import requests, os response = requests.post( 'https://api.interchange.io/api/v2/buyer/advertisers', headers={ 'Authorization': f'Bearer {os.environ["SCOPE3_API_KEY"]}', 'Content-Type': 'application/json', }, json={ 'name': 'Acme Corp (Sandbox)', 'description': 'Integration testing advertiser', 'sandbox': True, }, ) advertiser = response.json()['data'] print('Sandbox advertiser:', advertiser['id'], advertiser['sandbox']) ``` **Response:** ```json theme={null} { "data": { "id": "adv_abc123", "name": "Acme Corp (Sandbox)", "description": "Integration testing advertiser", "status": "active", "sandbox": true, "createdAt": "2026-02-22T21:35:44Z", "updatedAt": "2026-02-22T21:35:44Z", "linkedBrand": null, "brand": null, "brandWarning": null } } ``` ### Via UI When creating an advertiser in the dashboard, toggle the **Sandbox** switch before saving. Sandbox advertisers are shown with a badge in the advertiser list for easy identification. **Sandbox is permanent.** Once an advertiser is created with `sandbox: true`, the flag cannot be changed. This protects against accidentally switching an advertiser from sandbox to production after campaigns have been configured. *** ## Using Sandbox Once you have a sandbox advertiser, the entire workflow is identical to production. Discover products, create campaigns, add creatives, and execute — all using the same API endpoints. The sandbox routing is completely transparent. For example, executing a campaign: ```bash theme={null} curl -X POST "https://api.interchange.io/api/v2/buyer/campaigns/{campaignId}/execute" \ -H "Authorization: Bearer your-api-key" \ -H "Content-Type: application/json" ``` Because the campaign belongs to a sandbox advertiser, all ADCP calls are automatically routed through the sandbox environment. Responses are simulated — no real bids are placed and no real spend occurs. *** ## Murph Sandbox Test Plans When Murph is enabled for an account, `GET /api/v2/murph/test-runs` returns the latest sandbox test plans and diagnostics that appear in Murph's right rail. This endpoint is scoped to the authenticated account and requires the `ask-murph` feature flag. For a seller-owned inventory-source test, give Murph the advertiser's brand domain and currency during planning. Murph checks for a matching sandbox advertiser inside that account and returns the existing advertiser ID when one is available. The same brand-scoped sandbox account is used to preview products, so the selected product IDs remain valid when the test runs. If no matching advertiser exists, the read-only plan says that the confirmed execution will create one; it never substitutes a production advertiser. Sellers do not need buyer-side advertiser-list access or a previous test run to recover the ID. This is a buyer advertiser owned by the currently authenticated organization; it is not a hidden cross-organization identity or the seller-side advertiser inside the connected ad server. Query parameters: | Param | Notes | | -------------- | -------------------------------------------------------------------------------------------- | | `take` | Number of plans to return, 1..20. Defaults to 5. | | `storefrontId` | Optional storefront scope for accounts with more than one storefront. | | `runUid` | Optional exact test-run UUID. Use this when following a diagnostics link from a failed test. | Each item in `runs[]` includes the plan status (`planned`, `partial`, `passed`, or `failed`), the current stage, sanitized `summary`, `artifacts`, and `diagnostics`, plus timeline `steps[]` with operation names, HTTP status when available, and a diagnostic message. Murph currently records buyer-stack plans and approval-blocked attempts; executable buyer-stack runs require a server-issued approval token flow. When campaign execution fails, the tool result and stored run diagnostics distinguish local execution preflight from the buyer-side dispatch boundary. `executionAttempted: true` means the buyer stack evaluated a prepared media buy. `dispatchAttempted: false` means that evaluation failed before a `create_media_buy` dispatch began, so no seller call could have been issued. The associated `executionErrors[]` entry contains the media-buy ID, sales-agent ID, and specific preflight message. `dispatchAttempted: true` proves only that the buyer-side boundary was reached; an internal storefront can still fail before forwarding to its inventory source. `mediaBuysAttempted` counts prepared buys evaluated by `execute_campaign`; it does not by itself prove that a seller received a call. `diagnostics.agentCalls` lists the per-seller results returned by the discovery round-trip (`get_products`), including on runs that later fail at product selection or the dispatch boundary — a downstream failure does not clear the seller results discovery already returned. To confirm whether a seller was reached on the wire regardless of what discovery returned, cross-check the storefront activity log, which records the raw round-trip. These are safe summaries only, not raw upstream responses. When a Murph sandbox execution fails or stalls after recording a run, Murph opens Test Runs directly on that run. Hosts that cannot embed the MCP App, such as Slack, show an **Open diagnostics** button that opens the same exact-run view. This handoff is driven by the typed tool result and its `runUid`, not by matching words in Murph's answer. In an app-capable host, select **Fix this run** to have Murph diagnose the exact persisted failure and apply any safe repair available through typed tools. The result, any error, and **Try again** stay inside that run's detail; the action does not add a synthetic message to the host composer. If an external change or human decision is still required, the inline result names that prerequisite before you rerun the test. *** ## Filtering Sandbox Advertisers The `sandbox` field is returned on every advertiser response. Use the optional `sandbox` query parameter to filter: ```bash theme={null} # List only sandbox advertisers curl "https://api.interchange.io/api/v2/buyer/advertisers?sandbox=true" \ -H "Authorization: Bearer your-api-key" # List only production advertisers curl "https://api.interchange.io/api/v2/buyer/advertisers?sandbox=false" \ -H "Authorization: Bearer your-api-key" ``` In the dashboard, sandbox advertisers are shown with a **Sandbox** badge so they are easy to distinguish from production advertisers at a glance. *** ## Key Constraints | Constraint | Detail | | ---------------------- | ------------------------------------------------------------------------------------------------------------- | | **Immutable flag** | `sandbox` cannot be changed after advertiser creation | | **Strict isolation** | Sandbox and production data are fully isolated — no risk of mixing test and real traffic | | **No silent fallback** | If sandbox execution cannot proceed, the API returns an explicit error rather than falling back to production | *** ## Next Steps Full schema for `POST /advertisers`, including the `sandbox` field. Protocol-level details on how sandbox mode works in AdCP. Get up and running with the Interchange API.