Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
Meta Ads MCP Server Arrives in Databricks Marketplace - OpenSmartRoute
What Changed - The ads MCP server joins Databricks Marketplace
Databricks added Meta's ads Model Context Protocol server to its marketplace. This update arrived in October 2026. Marketers can now run campaigns directly with customer data inside Genie. They no longer need separate CSV exports or manual API integrations. The connection allows natural language exploration of campaign performance alongside governed business data. Administrators control access through Unity Catalog and Unity Gateway. These tools record usage and audit logs for every tool call.
The release expands the existing collaboration between Databricks and Meta. It follows the launch of Meta Conversions API on Databricks Marketplace earlier this year. That previous feature helped advertisers send first-party conversion signals from Databricks to Meta. The new server focuses on campaign lifecycle management instead. It exposes over 25 tools across the entire advertising workflow. These tools cover creating campaigns, changing budgets, and managing creatives.
The change bridges a gap between data teams and media teams. Previously, insights sat in one place while campaign decisions happened elsewhere. A churn model lived in production while budget adjustments occurred in Meta Ads Manager. The handoff required exporting data or scheduling recurring sync meetings. Now, the same conversation can access both the churn score and campaign results. This direct path removes friction from the decision-making process.
How It Works - Connecting Genie to Meta Tools via Natural Language
Model Context Protocol is an open standard that lets AI agents interact with tools and data sources. The ads MCP server from Meta exposes a suite of 25+ tools for marketers. These tools span the full campaign lifecycle from creation to diagnosis. Marketers can use Genie One to ask questions about campaign performance. They get instant insights grounded in enterprise context curated by data teams.
Genie One allows users to take action across external tools using natural language. Adding Meta's ads MCP server brings reporting and advertising tools into the same conversation. A marketer can prompt: "Use our churn-risk score to identify high-value customers at risk of leaving." Genie then searches for the appropriate data in Databricks. It finds relevant enterprise context across connected tools and systems. Finally, it creates a retention campaign with an $8,000 daily budget.
Simon Willison tested Claude Opus 5.5 on composing Monkey Island-style game music. The model produced surprisingly high-quality results in a text-based format.
NVIDIA's AVO agent system scored 100% on ARC-AGI-3 using Claude Opus 5. It completed all levels with fewer actions than the VISTA system.
Budget allocation becomes another area where agents can optimize spend. Gross margin per order lives in Databricks while campaign spend lives at Meta. An agent with access to both can recommend moving budget toward profitable revenue. It does not just look for cheap conversions anymore. The model understands how the business and marketing actually work together.
If advertisers are already sending conversions through the Conversions API, they gain new diagnostic capabilities. Agents can now investigate event volumes and Event Match Quality directly. They can ask if purchase events dropped because sales slowed or pipeline changed. This investigation draws on diagnostics from Meta and data in Databricks simultaneously. The Monday morning question about underperforming campaigns becomes one prompt instead of stitching two tools together.
Security and Control - Unity Catalog and Gateway Governance
Security remains a priority for enterprise advertisers using these new capabilities. Administrators control access to the Meta connection through Unity Catalog. Unity Gateway governs tool calls and records usage and audit logs. Enterprise data remains subject to existing Databricks permissions. Actions in Ads Manager also follow strict governance rules.
At Meta, advertisers set specific rules in Business Settings. They can block any budget increase over 20%. They can disallow campaign creation on a given account entirely. The server checks every tool call against these rules before it executes. If an agent breaks one rule, the call is blocked immediately. A structured error back to the agent explains why the action failed. This is enforced server-side so it holds regardless of which model you point at it.
You can manage the same rules through the Marketing API. This matters if you are running hundreds of ad accounts manually. You do not need to click through settings pages one by one. The system ensures that no unauthorized changes happen to production ad accounts. Safety is built into the architecture rather than relying on prompt engineering.
The 25+ Tools Available for Campaign Lifecycle Management
The ads MCP server provides a comprehensive set of tools for campaign management. It covers creating campaigns and ad sets within Meta's platform. Marketers can change budgets and bids dynamically through natural language prompts. Defining audiences becomes part of the same conversation as data analysis. Managing creative assets and catalogs is also supported by these tools.
Pulling performance data is another core function of the server. Diagnosing signal quality helps maintain data integrity for future campaigns. The list includes over 25 distinct tools across the lifecycle. This breadth allows agents to handle end-to-end workflows without human intervention. Each tool integrates with the broader Databricks ecosystem for context.
Why it Matters - Bridging Data Teams and Media Teams
Most integrations let an agent manage ads in isolation. In Databricks, an agent can manage ads while reasoning over your whole data estate. It considers segments, propensity scores, LTV, first-party attribution, and inventory forecasts. This is the difference between automating a few clicks and making a better call than the workflow it replaced.
Customers want a more direct way to put their data and AI investments to work. For advertisers, this means putting customer intelligence at the center of decisions about audiences, offers, and budgets. The partnership with Meta helps customers connect the data and models they built in Databricks to campaign execution. They get the governance and control they need to scale these operations effectively.
Stephen Orban, SVP Product Partnerships and Ecosystem at Databricks, highlighted this focus on customer intelligence. He noted that teams have already invested in understanding their customers deeply. This gives them a more direct way to put that understanding to work in Meta campaigns. If you have models in Databricks and a media team working off platform reporting, this is the shortest path between having data and seeing it reflected in campaigns.
How it Compares - What Existed Before and What Stays the Same
Before this release, marketers used separate tools for data and ads. They exported CSV files or used custom APIs. This created a handoff between teams. The media team adjusted budgets in Meta Ads Manager manually. The data team updated models in Databricks separately. Insights often sat unused until a manual sync occurred.
The new ads MCP server changes how agents access tools. It connects directly to Meta campaign performance data. Marketers can now ask natural language questions about both datasets. They do not need to export files or maintain custom integrations anymore. The server exposes over 25 tools across the campaign lifecycle. These include creating campaigns, changing bids, and managing creatives.
What stays the same is the core requirement for permissions. Administrators still control access through Unity Catalog. Enterprise data remains subject to existing Databricks permissions. Actions in Ads Manager still require proper authorization. The server checks tool calls against Meta Business Settings rules. It blocks budget increases over 20% if configured. It prevents campaign creation on specific accounts entirely.
The integration replaces manual CSV exports and API maintenance. It removes the need for recurring Thursday sync meetings. Agents can now diagnose signal quality automatically. They can check event volumes and data freshness directly. The workflow moves from stitching tools together to one prompt. This reduces the time between insight and action significantly.
Questions This Leaves Open - What the Source Does Not Say and How a Reader Can Check It
The source text does not specify the exact cost of using the ads MCP server. Readers must check Databricks Marketplace for pricing details. There is no mention of specific latency figures for tool calls. Performance depends on network conditions between Databricks and Meta servers.
Readers should verify how many concurrent agents can use the server. The text mentions hundreds of ad accounts but does not give a limit. They need to test their own environment before scaling usage. Checking the Marketplace documentation is the first step here. Reading the API reference will reveal rate limits and quotas.
The source does not detail how often Meta updates its tool definitions. Marketers must monitor for new tools or deprecated features. A version mismatch could break existing agent workflows unexpectedly. Testing with a small budget increase helps confirm rule enforcement. Monitoring Unity Gateway logs confirms all actions are recorded properly.
Readers should ask if the server supports custom event tracking beyond Conversions API. The text mentions signal diagnostics but not full event schema flexibility. Custom integrations might still be needed for niche data types. Checking Meta's developer documentation clarifies supported event structures.
There is no information on how the server handles regional data restrictions. Meta and Databricks operate in different geographic regions. Data residency laws could impact cross-border campaign optimizations. Readers must ensure their ad accounts comply with local regulations.
The source does not state whether the server supports legacy API versions. Older marketing platforms might use deprecated endpoints. Agents could fail if the server only speaks modern protocols. Checking compatibility notes in the Marketplace listing is essential. Testing with a sandbox account avoids production risks.
Readers should verify how the server handles multi-language campaigns. Meta Ads Manager supports many languages for creative assets. The agent's ability to understand non-English data remains unclear. Natural language prompts might fail if the model lacks translation context.
There is no mention of how the server integrates with third-party ad networks. It currently focuses on Meta platforms specifically. Expanding to Google or Amazon requires separate integrations. Marketers building a unified media stack will need multiple servers.
The source does not explain how often agents refresh campaign data. Real-time updates could improve decision speed significantly. Batch processing might introduce delays in budget adjustments. Checking the server's polling frequency helps estimate response times.
Readers should confirm if the server supports automated A/B testing workflows. The text mentions diagnosing underperformance but not automatic test creation. Agents might need human approval for new experiments. This limits the degree of automation possible today.
The source does not detail how the server handles budget overruns. Meta blocks increases over 20% by default rules. Custom rules might allow higher limits with approval workflows. Readers must configure these thresholds carefully in Business Settings.
There is no information on how the server interacts with programmatic buying. It focuses on manual campaign management tools mostly. Programmatic exchanges require different API structures entirely. Marketers using DSPs will need additional setup steps.
The source does not state whether the server supports historical data analysis. It pulls current performance metrics primarily. Analyzing past trends requires querying Databricks directly first. The agent might lack context for long-term pattern recognition.
Readers should verify how the server handles negative placements. Meta allows blocking specific audiences or creatives automatically. Agents can execute these blocks via tool calls easily. This feature reduces wasted spend on bad targeting options effectively.
The source does not explain how the server manages creative approval workflows. Meta requires human review for certain ad types before publishing. The agent cannot bypass this regulatory requirement entirely. Marketers must still approve final creative assets manually.
There is no mention of how the server handles currency conversions. Campaign budgets and revenue data use different currencies often. The agent might struggle with accurate financial calculations across borders. Readers need to ensure consistent unit definitions in Databricks.
The source does not detail how the server interacts with Meta's new AI features. Meta is rolling out generative AI for ad creation. The server exposes tools but does not generate creative content itself. Marketers still need human input for final creative decisions.
Readers should check if the server supports real-time bid optimization. It can change bids based on performance data quickly. However, automated bidding strategies require Meta's specific API permissions. The server acts as a bridge, not a full replacement.
The source does not state how long it takes to provision a new connection. Setup involves creating user access tokens and configuring Unity Catalog. This process might take hours depending on administrative approval speed. Planning ahead prevents workflow disruptions during peak seasons.
There is no information on how the server handles data privacy compliance. GDPR and CCPA regulations apply to customer data usage. The agent must respect these legal constraints when accessing personal info. Readers need to ensure their data policies align with tool usage.
The source does not mention whether the server supports offline campaign modes. Some regions or events might lack internet connectivity for real-time updates. Agents could fail to execute bids during network outages temporarily. Backup plans are necessary for critical campaigns.
Readers should verify how the server handles multi-account management. It supports hundreds of ad accounts through centralized rules. Managing thousands of accounts manually is impossible without this tool. This feature scales operations significantly for large advertisers.
The source does not explain how the server integrates with Meta's Advantage+ tools. These automated tools use AI to optimize campaigns independently. The MCP server complements but does not replace these systems entirely. Marketers might run both strategies simultaneously.
There is no mention of how the server handles creative asset storage. It pulls performance data but does not host creative files directly. Media teams must store creatives in approved Meta locations. The agent references URLs rather than hosting content itself.
The source does not state whether the server supports international audience targeting. Meta allows targeting by country, language, and interest globally. The agent can query this data from Databricks easily. However, local ad laws might restrict certain demographic data access.
Readers should check if the server supports retargeting workflows specifically. It can identify churn-risk customers and launch retention campaigns. This is a specific use case highlighted in the announcement. Implementing this requires careful segmentation logic in Databricks.
The source does not explain how the server handles campaign fatigue detection. Meta provides signals for creative performance decay over time. The agent can recommend budget shifts to new creatives automatically. This feature improves long-term campaign efficiency significantly.
There is no information on how the server integrates with Meta's Pixel events. It uses Conversions API and signal diagnostics primarily. Direct Pixel integration might require additional configuration steps. Marketers should verify event mapping accuracy in their setup.
The source does not mention whether the server supports multi-language reporting. Campaign performance data exists in multiple languages often. The agent translates findings into natural language prompts effectively. This helps non-English speaking stakeholders understand results better.
Readers should verify how the server handles cross-channel attribution. Meta focuses on its own platform data mostly. Databricks provides first-party attribution models separately. Combining these sources gives a complete view of customer journeys.
The source does not state if the server supports real-time bidding adjustments. It can change bids based on performance metrics quickly. However, programmatic real-time bidding requires different API access levels. Marketers using DSPs need separate integrations for this feature.
There is no mention of how the server handles negative placement rules. Meta allows blocking specific audiences or creatives automatically. The agent can execute these blocks via tool calls easily. This reduces wasted spend on bad targeting options effectively.
The source does not explain how the server manages creative approval workflows. Meta requires human review for certain ad types before publishing. The agent cannot bypass this regulatory requirement entirely. Marketers must still approve final creative assets manually.
There is no information on how the server handles currency conversions. Campaign budgets and revenue data use different currencies often. The agent might struggle with accurate financial calculations across borders. Readers need to ensure consistent unit definitions in Databricks.
The source does not detail how the server interacts with Meta's new AI features. Meta is rolling out generative AI for ad creation. The server exposes tools but does not generate creative content itself. Marketers still need human input for final creative decisions.
Readers should check if the server supports real-time bid optimization. It can change bids based on performance data quickly. However, automated bidding strategies require Meta's specific API permissions. The server acts as a bridge, not a full replacement.
The source does not state how long it takes to provision a new connection. Setup involves creating user access tokens and configuring Unity Catalog. This process might take hours depending on administrative approval speed. Planning ahead prevents workflow disruptions during peak seasons.
There is no information on how the server handles data privacy compliance. GDPR and CCPA regulations apply to customer data usage. The agent must respect these legal constraints when accessing personal info. Readers need to ensure their data policies align with tool usage.
The source does not mention whether the server supports offline campaign modes. Some regions or events might lack internet connectivity for real-time updates. Agents could fail to execute bids during network outages temporarily. Backup plans are necessary for critical campaigns.
What to Do - Getting Started with the New Server
To get started, find the ads MCP server from Meta in the Databricks Marketplace. Follow the instructions to configure a connection using a Meta user access token. You will need the required permissions for your specific ad accounts. Give your agent access to the relevant marketing data in Databricks as well. Begin with a concrete question your team already asks every week without hesitation.
Your teams have already invested in understanding their customers thoroughly. This investment gives them a more direct way to put that understanding to work in Meta campaigns. If you have models in Databricks and a media team working off platform reporting, this is the shortest path between "we have the data" and "the campaign reflects it."
Check the Marketplace documentation for the latest version of the server. Verify your Unity Catalog permissions before activating the connection. Test the agent with a small budget increase to ensure rule enforcement works correctly. Monitor the audit logs in Unity Gateway to confirm all actions are recorded properly.
