Show HN: Airbyte Agents – context for agents across multiple data sources

The Hidden Bottleneck Killing AI Agents: Why Context Is the New Frontier of Automation

In the race to build intelligent AI agents that can autonomously manage workflows, schedule meetings, close deals, and resolve customer issues, one critical component has been consistently overlooked: context. While much of the AI world has focused on improving reasoning, language understanding, and tool-calling capabilities, the real bottleneck lies not in the brain of the agent—but in its access to the right data, at the right time, in the right format.

Enter Airbyte Agents, a groundbreaking new framework that rethinks how AI systems interact with operational data. Rather than forcing agents to navigate a maze of APIs, authentication layers, and fragmented schemas every time they need information, Airbyte Agents introduces a unified context layer—a smart data index optimized specifically for agentic workflows. This isn’t just another API wrapper. It’s a fundamental shift in how we architect systems for autonomous intelligence.

For six years, Airbyte has been quietly building the world’s most robust library of data connectors—tools that sync data from sources like Salesforce, Slack, GitHub, and Zendesk into warehouses. Now, the company is repurposing that infrastructure for a new era: one where AI agents don’t just read data, but understand it across systems.

The Problem with Today’s AI Agents: They’re Blind Without Context

Imagine asking an AI agent to determine which enterprise customers are at risk of churning this quarter. On the surface, it sounds simple. But the reality is a labyrinth of disconnected systems.

The agent must first identify accounts in Salesforce, then cross-reference them with support ticket volumes in Zendesk, check for recent product usage drops in Amplitude, verify if a customer success manager has logged a risk flag in Notion, and possibly even scan Slack channels for internal warnings. Each step requires a different API call, often with unique authentication, pagination logic, and schema quirks.

Worse, most agents today are built on top of Model Context Protocols (MCPs)—thin wrappers around APIs. These tools assume the agent already knows what to query. But in real-world scenarios, agents rarely start with a known endpoint or object ID. They begin with a high-level question and must discover the relevant data first.

This leads to what Michel, co-founder and CEO of Airbyte, calls the “47-step trace”: a chaotic sequence of API calls, entity mappings, and error-prone joins that slow down responses and degrade accuracy. In one real-world test, an agent took over a minute and nearly fifty steps to answer a seemingly straightforward question—only to deliver a wrong answer.

The Birth of the Context Store: A Data Index for Agents

To solve this, Airbyte built the Context Store—a centralized, agent-optimized data index that sits between AI systems and operational tools. Unlike traditional data warehouses, which are built for analytics and batch processing, the Context Store is designed for real-time, agent-driven discovery.

Think of it as a “Google for your company’s operational data.” But instead of indexing web pages, it indexes entities—customers, deals, tickets, issues—across systems, preserving relationships and context. When an agent asks, “Which enterprise deals closing this month have open support tickets?” the Context Store doesn’t just return raw records. It returns a structured, cross-system view: Deal X in Salesforce is linked to Ticket Y in Linear, which references GitHub Issue Z.

This is made possible by Airbyte’s six years of work on data replication. Every connector—whether for HubSpot, Intercom, or Jira—has been refined to not only move data but to enrich it with metadata, lineage, and semantic context. That foundation now powers the Context Store, enabling agents to reason over a unified data graph.

Why APIs Alone Can’t Power Autonomous Agents

APIs were designed for human developers, not autonomous systems. They assume prior knowledge: you know the endpoint, the object ID, the field names. But agents operate differently. They start with intent, not instructions.

For example, when an agent is asked to “find every support ticket without a linked GitHub issue,” it doesn’t know which tickets or issues to look for. It must first discover the relevant datasets, understand their schemas, map entities across systems, and then perform the join—all in real time.

This is where traditional APIs fail. They offer no semantic layer. No way to say, “Show me all unresolved customer issues.” Instead, you must manually construct queries, handle pagination, manage rate limits, and stitch together results.

Airbyte Agents solves this by introducing a semantic query layer. Agents can now ask questions in natural language or structured intent, and the system translates them into optimized, cross-system queries. The Context Store handles the complexity behind the scenes—resolving entities, joining data, and returning actionable insights.

Real-World Impact: From Chaos to Clarity

Let’s revisit the original example: identifying at-risk customers. With traditional tools, the agent’s trace looked like a Rube Goldberg machine—47 steps of API calls, data parsing, and error handling. The final answer was not only slow but often incorrect due to stale data or missed connections.

With Airbyte Agents, the process is radically simplified. The agent queries the Context Store with a single intent: “Find customers with declining usage, high ticket volume, and no recent CSM touchpoint.” The system returns a ranked list of at-risk accounts, enriched with context from Salesforce, Zendesk, and product analytics.

This isn’t just faster—it’s more reliable. Because the Context Store is continuously synced, agents work with up-to-date information. And because relationships are pre-resolved, there’s no risk of mismatched entities or orphaned records.

Early adopters report dramatic improvements. One SaaS company reduced its churn prediction time from hours to minutes. Another automated its quarterly business review process, with agents pulling data from 12 different tools into a single report.

The Future of Agentic Workflows: Context as a Service

Airbyte Agents represents a shift toward Context as a Service (CaaS)—a new paradigm where context isn’t something agents must build on the fly, but a foundational layer they can rely on.

This has profound implications. As AI agents take on more complex roles—from sales automation to IT support to financial forecasting—they’ll need access to rich, cross-system context. The ability to understand not just what data exists, but how it relates to other data, will separate intelligent agents from mere automation scripts.

Moreover, the Context Store opens the door to agent collaboration. Imagine multiple agents working together: one analyzing customer sentiment in Slack, another tracking deal progress in Salesforce, and a third monitoring system health in Datadog. With a shared context layer, they can coordinate, share insights, and act in concert—without duplicating effort or stepping on each other’s toes.

Building the Benchmark: Proving the Value of Context

When Michel first saw the 47-step agent trace, he knew something had to change. But to convince the world, he needed data—not just anecdotes.

Over a single weekend, he built a benchmark harness to test Airbyte Agents against traditional API-based approaches. The results were staggering. Not only did the Context Store reduce step count and latency, but it also improved accuracy by nearly 40 percentage points.

The benchmark tested five common agentic workflows:

• Churn risk prediction
• Deal pipeline analysis
• Support ticket triage
• Cross-system issue tracking
• Automated reporting

In every case, Airbyte Agents outperformed the control group. The biggest gains came in tasks requiring entity resolution and cross-system joins—precisely the areas where APIs fall short.

This wasn’t just a win for Airbyte. It was validation of a broader thesis: context is the missing ingredient in agentic AI.

The Road Ahead: From Data Connectors to Intelligent Infrastructure

Airbyte’s journey—from data replication to agentic context—mirrors the evolution of enterprise AI itself. First, we connected systems. Then, we analyzed data. Now, we’re enabling intelligent action.

The launch of Airbyte Agents marks a turning point. It’s no longer enough to move data. We must make it understandable—to machines as well as humans.

As AI agents become embedded in workflows across sales, support, engineering, and operations, the demand for unified context will only grow. Companies that invest in context layers today will gain a decisive advantage: faster decisions, smarter automation, and more reliable AI.

The future of work isn’t just about smarter models. It’s about smarter systems—systems where context isn’t an afterthought, but the foundation.

And with Airbyte Agents, that future is already here.

This article was curated from Show HN: Airbyte Agents – context for agents across multiple data sources via Hacker News (Top)