Stacks / 8 workflow stages · v3.1.0 · reviewed 2026-09-05
Build a Production Research Agent
From web discovery to extraction, orchestration, browser actions, tracing, evaluation, and security.
Best for
Teams building an agent that has to find information, inspect sources, take actions, and produce evidence-backed output reliably.
Outcome
A research agent that can discover sources, retrieve usable content, navigate the web when APIs are not enough, and be traced and tested before production.
Not for
A simple one-shot search or chat experience that does not need tools, state, browser actions, or production monitoring.
Operator guidance
Best whenThe agent must discover, retrieve, act, and verify across a multi-step research workflow.
Avoid whenA single retrieval call or analyst-led workflow is sufficient.
CompatibilityPass structured citations and trace IDs between retrieval, orchestration, and evaluation; keep browser actions behind explicit permissions.
Use this as a decision framework, not a mandatory shopping list. Swap or skip layers based on your existing stack, constraints, and risk profile.
1. Discover
Use a search layer designed for programmatic and semantic discovery so the agent can find relevant sources instead of relying on a fixed corpus.
- Tavily
Recommendation details
Why: Consider Tavily as an alternative for 1. Discover; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Search API built for AI agents.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Tavily official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- Brave Search API
Recommendation details
Why: Consider Brave Search API as an alternative for 1. Discover; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Independent web search API.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Brave Search API official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Search and retrieval API useful for Claude agents.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Exa official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
2. Retrieve
Turn web pages into cleaner machine-usable content for downstream reasoning and extraction.
- Jina AI Reader
Recommendation details
Why: Consider Jina AI Reader as an alternative for 2. Retrieve; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Convert web pages into LLM-friendly text.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Jina AI Reader official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- Apify
Recommendation details
Why: Consider Apify as an alternative for 2. Retrieve; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Large automation/scraping actor ecosystem.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Apify official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Web extraction and crawling for agent workflows.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Firecrawl official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
3. Structure
Give the agent typed tools and structured outputs so important steps are less dependent on free-form model text.
- Claude Agent SDK
Recommendation details
Why: Consider Claude Agent SDK as an alternative for 3. Structure; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Programmatic agent workflows using Claude Code capabilities.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Claude Agent SDK official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Typed agent framework from Pydantic.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Pydantic AI official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
4. Orchestrate
Add explicit state and multi-step control when the workflow needs branching, retries, checkpoints, or durable execution.
- LlamaIndex
Recommendation details
Why: Consider LlamaIndex as an alternative for 4. Orchestrate; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Retrieval and agent framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: LlamaIndex official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- Dify
Recommendation details
Why: Consider Dify as an alternative for 4. Orchestrate; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Production-ready platform for agentic workflows, RAG, model management and observability.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Dify official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Stateful agent/workflow framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: LangGraph official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
5. Act in the browser
Use managed browser infrastructure when the agent has to interact with real web interfaces rather than only fetch pages.
- Browser Use
Recommendation details
Why: Consider Browser Use as an alternative for 5. Act in the browser; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Open-source browser harness for AI agents, with Claude Code skill support and cloud/browser infrastructure.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Browser Use official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- Stagehand
Recommendation details
Why: Consider Stagehand as an alternative for 5. Act in the browser; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: AI-oriented browser automation framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Stagehand official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- Playwright
Recommendation details
Why: Consider Playwright as an alternative for 5. Act in the browser; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Reliable browser automation Claude can operate through code.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Playwright official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Managed browsers for agents.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Browserbase official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
6. Trace
Capture prompts, tool calls, traces, latency, and failures so you can see what the agent actually did.
- LangSmith
Recommendation details
Why: Consider LangSmith as an alternative for 6. Trace; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Tracing and observability platform for agent and LLM applications.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: LangSmith official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- Arize Phoenix
Recommendation details
Why: Consider Arize Phoenix as an alternative for 6. Trace; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Open-source tracing and evaluation.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Arize Phoenix official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Open-source LLM observability and evals.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Langfuse official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
7. Evaluate
Create repeatable tests for prompts, tool behavior, and regressions before changes reach users.
- Braintrust
Recommendation details
Why: Consider Braintrust as an alternative for 7. Evaluate; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Evals, traces and production quality workflows.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Braintrust official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- DeepEval
Recommendation details
Why: Consider DeepEval as an alternative for 7. Evaluate; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Code-first evaluation framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: DeepEval official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Open-source LLM eval and red-team framework.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: Promptfoo official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
8. Harden
Review common GenAI and agent attack surfaces before exposing tools, data, or browser actions in production.
- MITRE ATLAS
Recommendation details
Why: Consider MITRE ATLAS as an alternative for 8. Harden; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: Adversarial threat knowledge base for AI.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: MITRE ATLAS official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
- NVIDIA garak
Recommendation details
Why: Consider NVIDIA garak as an alternative for 8. Harden; compare its documented capabilities and operating constraints with the primary recommendation.
Best when: LLM vulnerability scanner.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: NVIDIA garak official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Recommendation details
Best when: Core security guidance for GenAI systems.
Avoid when: Avoid when the current official source does not support your required deployment, data-handling, or integration constraints.
Compatibility: Validate interfaces, data formats, and operational handoffs with the adjacent stack stages.
Prerequisites:
- Review the current official source and implementation requirements.
Evidence: OWASP GenAI Security Project official source (2026-09-05)
Verified 2026-09-05 · Medium confidence
Implementation references
-
Anthropic Agent Patterns Cookbook ↗
Cookbook implementations of the workflow and agent patterns described in Anthropic's Building Effective Agents guidance.
github.com · cookbook -
OpenAI Agents Cookbook ↗
Current implementation examples for agent workflows, tools, traces, and evaluation loops.
developers.openai.com · cookbook
Need this workflow built and running, not just linked? EE Solutions implements agentic stacks like this one. EE Solutions is a senior technology team for private capital firms and their portfolio companies.
Talk to EE Solutions ↗