Today, we're launching Agent Composer—new capabilities for our platform that reduce the time needed to complete complex technical tasks from hours to minutes, and free engineering teams to focus on highest-impact problems.
Teams can now build and customize dynamic agents and static workflows using an intuitive development experience, comprehensive AI toolkit, and enterprise-scale runtime.
Agent Composer extends our unified context layer with orchestration capabilities that enable multi-tool coordination, multi-step reasoning, and complex workflows across enterprise data.
Complex technical work has been neglected by AI
Generic AI has transformed how work is done, but it still fails at complex, specialized work. Engineering in semiconductors, electronics, aerospace, manufacturing, and logistics requires precision that LLM models alone can't deliver.
This means that while highly trained knowledge workers can use AI to draft emails or summarize meetings, complex engineering tasks still consume dozens of hours per week (e.g. analyzing gigabytes of test data, synthesizing technical documentation, generating production plans).
Without domain-specific AI, a rocket propulsion test engineer's typical week is literally rocket science—which actually means hours of complex, routine work:
4 hours: Review overnight hot-fire test results for anomalies against historical baselines
- A single 30-second engine firing generates gigabytes of telemetry across hundreds of sensor channels—chamber pressure, turbopump RPM, injector temperatures, thrust measurements.
4 hours: Answer complex technical questions during anomaly investigation: "What could cause high-frequency combustion instability in a pintle injector at this mixture ratio?"
- Synthesizing answers requires searching engine design docs, propellant specifications, NASA technical standards, AIAA guidelines, and historical anomaly reports.
8 hours: Writing and debugging test control code for the next campaign.
- The code must implement procedures while enforcing safety interlocks, mapping to DAQ channels and valve drivers, and validating against prior scripts.
10 hours: Assemble and cross validate the Test Readiness Review package.
- Hunt down dozens of documents—approved procedures, hazard analyses, calibration records, personnel certifications, facility permits—and validate everything is current.
20+ hours spent on critical but routine knowledge work, leaves limited time for design improvements or novel R&D.
Rocket science makes the problem vivid, but it's not unique. Engineers at R&D-focused companies across industries lose dozens of hours weekly to root causing errors, generating production plans, writing test code, and validating compliance without AI assistance.
The real barrier: solving the context problem
The lack of expert AI is not a model problem, it's a context problem. Enterprises struggle to connect powerful models to their messy, specialized data and institutional knowledge.
Solving this requires context engineering: the discipline of building expert AI that reliably automates complex, repetitive tasks. When enterprises try to solve this, they face a hard tradeoff.
Build it yourself: Fully customizable, but requires scarce AI expertise, months of development, and constant optimization, with no guarantee of project success
Buy off-the-shelf: Fast to deploy, but inflexible. Hard to customize and doesn't scale across use cases.
The third way: a platform approach
Unlike DIY and point solutions, we offer a unified context layer specialized for domain-specific tasks in technical industries. This enables enterprises to deliver expert AI apps easily adapted to their needs in weeks, not months.
How Agent Composer makes expert AI agents accessible
Today we're adding Agent Composer to our context layer platform. It’s a new infrastructure and orchestration layer that manages context throughout multi-step agent workflows, enforces guardrails, and maintains agent reliability.
The Agent Composer has three main components:
- Agent runtime: Run agents at enterprise scale without losing uptime or accuracy
- Intuitive development experience: Build agents without writing a single line of code with prebuilt agents, agent prompt builder and a GUI drag and drop experience.
- Comprehensive library of AI tools: Ensure agents are fully immersed in your enterprises’ data and documents with accuracy optimized tools for planning, retrieval, ingestion, taking actions, evaluation, and managing memory.
The new capabilities unlocked by dynamically orchestrating AI tools
These development capabilities are powered by upgraded orchestration that enables workflows previously impossible with static retrieval:
- Multi-step reasoning—Decompose problems, iterate solutions, revise outputs. Move beyond single-hop Q&A.
- Multi-tool orchestration—Coordinate across docs, logs, web search, and APIs. Handle structured and unstructured data in the same workflow.
- Customizable workflows and outputs—Generate ready-to-use reports, plans, investigations with clear sections and artifacts tailored to specifications.
- Hybrid agentic behavior—Combine dynamic agent steps with static workflow control for the right balance of intelligence and safety.
- One-click agent optimzation—Automatically optimize agent performance based on user feedback.
Let's explore how Agent Composer makes building agents easier and unlocks capabilities for use cases previously too complex for AI.
Three ways to build an agent
Agent Composer makes developing even the most complex customized agents simple through three development paths. Teams can combine dynamic orchestration with static workflows in the same agent. This provides flexibility for ad-hoc requests while maintaining control over mission-critical steps.
Start with a pre-built agent: Use one of our pre-built agents to get started, and jump right into specialization and agent tuning. Teams can also save and share their own agents to scale their impact across the business.
| Basic Search | Quickly generate accurate answers to straightforward questions by retrieving context from large volumes of enterprise data sources. |
| Agentic Search | Solve complex problems end-to-end using multi-step reasoning, tool coordination, and stateful context. |
| Deep Research | Produce comprehensive research outputs by synthesizing information across documents, logs, and web sources. |
| Root Cause Analysis | Identify root causes by retrieving from diverse data sources and iteratively validating hypotheses through evidence-based reasoning. |
| Task Execution | Automate repetitive business processes using agents that execute API write actions across enterprise systems. |
| Structured Extraction | Reliably extract complex structured data into clean JSON. Quickly verify with groundedness scores, and page-level attribution |
- Build from a prompt: Best for business users and rapid prototyping. Create an agent by describing your use case or workflow in natural language. In a few moments you will have a working agent to query or refine.
- Build from a blank canvas: A visual drag-and-drop development experience, best for use cases that require custom logic, specialized integrations, or novel workflows. Developers have complete control with every step the agent takes.
Accuracy and agent specialization tools
Building an agent takes only a few minutes, but production readiness is equally fast. Agent Composer makes it easy to specialize agents to specific use cases and data types. Every node of an agent is completely customizable, enabling precise tuning of prompts, retrieval parameters, and tool selection. Unlike DIY offerings, Agent Composer is completely model agnostic, ensuring you aren't locked in and can select the best model for the job.
Agent accuracy can also be automatically optimized with a single click inside Agent Composer. In combination with the Feedback Annotation feature, the agent optimizer uses feedback from previous query to optimize prompts and other agent configs with no human intervention.
The impact: Giving engineers their week back
These capabilities fundamentally transform how engineers work. That rocket propulsion engineer spending 20-26 hours on routine tasks? Expert AI agents and workflows compress that work into a fraction of the time, freeing engineers to focus on high-value analysis and innovation.
| Job to be done | Agent workflow | Before | After |
|---|---|---|---|
| Test telemetry analysis | Test data anomaly detection | 4 hours | 20 minutes |
| Technical question answering | Technical issue resolution | 4 hours | 10 minutes |
| Test code creation | Test code generation | 4-8 hours | 30-60 minutes |
| TRR package assembly | TRR package preparation | 8-10 hours | 1-2 hours |
Each workflow accesses the same data sources the engineer would manually search—telemetry files, design docs, safety standards, historical data, equipment specs—but synthesizes information in minutes instead of hours. The engineer gets their week back to do what only humans can: make critical design decisions, explore novel approaches, push R&D forward.
Beyond rocket science: Solving complex knowledge work across industries
Agent Composer's orchestration capabilities, multi-step reasoning, multi-tool coordination, stateful context, enable expanded use cases across technical disciplines.
Early customers are seeing this impact across industries:
- An advanced manufacturer reduced root-cause analysis from 8 hours to 20 minutes by automating sensor data parsing and log correlation.
- A global strategy consulting firm reduced manual research from hours to seconds, giving consultants access to relevant case work, answers to complex questions, and prior examples.
- A tech-enabled 3PL provider achieved 60x faster issue resolution by providing instant answers across their entire internal knowledge base.
- A specialty chemicals manufacturer reduced product research from hours to minutes with agents that search patents and regulatory databases.
- A test equipment maker generates test code in minutes instead of days by translating procedures into control logic.
Get started
Agent Composer delivers 10x productivity gains by solving complex, routine workflows that generic AI can't handle. We combine platform capabilities, AI expertise, and engineering experience to build specialized AI that works on messy enterprise data.
- Join our webinar on 2/5 to get more technical details, directly from our founder and CEO
- Experience the upgraded capabilities by playing with our demo agent app.
- Try building an agentic search agent yourself in our free trial. Utilize global datastores to get started in seconds.
- Contact us to learn how Agent Composer can transform your most time-intensive workflows.
