# AI agent development tools for developers

Compare tools for retrieval, browser automation, agent runtimes, and in-app AI using Console’s dated reviews. Start with the capability your application needs, then check integration approaches and tradeoffs.

Source: https://console.dev/tools/categories/ai-agent-development

Reviews by: [David Mytton](https://console.dev/about#david-mytton)

Comparison updated: 2026-10-17 (original review dates below)

Console reviews published in 2024–2026. The table links to each tool’s latest review; excerpts below retain their original review dates. Features, pricing, and limitations may have changed. This is a starting shortlist, not a current product ranking.

## How we selected these

Selected reviews covering retrieval, browser actions, agent tooling, workflow composition, and user interfaces. Listed alphabetically. These are different building blocks, not a tested integrated stack or a ranking.

[Our review criteria](https://console.dev/selection-criteria)

## Find your starting point

| Tool | Consider it for | Integration described | Latest review |
| --- | --- | --- | --- |
| [Atomic](https://console.dev/tools/atomic) | A local Markdown knowledge base with semantic search | Self-hosted knowledge base; MCP or HTTP server | [2026-04-09](https://console.dev/tools/atomic#review_65e803d23319617e1fb8950b69fbbca683bed8868d466faaf6048088f0ac8a7b) |
| [Exa](https://console.dev/tools/exa) | Web search for LLM workflows | Search API; Python, Go and Node SDKs at review | [2024-05-30](https://console.dev/tools/exa#review_2bf011eae5003f739ea126017973105b9b13380e52a8da089d48fb93ae11ff53) |
| [Go Micro](https://console.dev/tools/go-micro) | Agent runtimes and service orchestration in Go | Go interfaces; MCP and A2A gateways | [2026-06-25](https://console.dev/tools/go-micro#review_1006c6f6e2df9304689388312390f284e9ecb5c9901e9e75b4e9662f3b1eeca1) |
| [Langbase](https://console.dev/tools/langbase) | Composing LLM workflows and managed retrieval | Composable pipes; multiple LLM APIs and managed RAG | [2024-08-08](https://console.dev/tools/langbase#review_3d231cb5ecc0551bd891aea24e76aff8bd8787c4c81612e06f6891c9d0320835) |
| [Liveblocks AI Copilots](https://console.dev/tools/liveblocks-ai-copilots) | Embedding AI interactions in a React app | React components; custom tools and your own LLM backend | [2025-06-19](https://console.dev/tools/liveblocks-ai-copilots#review_af306e62d8030cb787207b90f95876c099aa1d15241f676b51722d98dc2212f4) |
| [Mu](https://console.dev/tools/mu) | Ready-made tools for agents | MCP server, browser UI and CLI; self-hosted or cloud | [2026-08-06](https://console.dev/tools/mu#review_9763dbcf2998584c4ff9ac18027093e95bb982c3911e11af315f838b63371de8) |
| [qmd](https://console.dev/tools/qmd) | Searching local Markdown from an agent | On-device CLI and MCP server | [2026-01-22](https://console.dev/tools/qmd#review_d762ccd839607b3324757453e45200891d8d20e6d9bbd5a50376ae6c4628b674) |
| [Stagehand](https://console.dev/tools/stagehand) | Browser automation that combines code and AI | Playwright-compatible commands with AI fallbacks | [2025-01-16](https://console.dev/tools/stagehand#review_a030d6868bd8c4b5856ed88333f0e6824183c00ea719d975400ddc14ba7db0f3) |

## The reviews behind the comparison

### Atomic

Consider for: A local Markdown knowledge base with semantic search

Integration: Self-hosted knowledge base; MCP or HTTP server

Reviewed: 2026-04-09

Review: https://console.dev/tools/atomic#review_65e803d23319617e1fb8950b69fbbca683bed8868d466faaf6048088f0ac8a7b

#### What we liked

Atomic — Console review, Apr 9, 2026. Local, open source, markdown-based knowledge base. Auto-generates wiki articles from tags. Built-in semantic search with vector embeddings. Supports different types e.g. markdown notes, web clips, articles. Supports MCP. Can be an HTTP server without a desktop UI.

#### Tradeoffs at review

Atomic — Console review, Apr 9, 2026. Everything is self hosted so you need to do some setup work. This is also a benefit.

### Exa

Consider for: Web search for LLM workflows

Integration: Search API; Python, Go and Node SDKs at review

Reviewed: 2024-05-30

Review: https://console.dev/tools/exa#review_2bf011eae5003f739ea126017973105b9b13380e52a8da089d48fb93ae11ff53

#### What we liked

Exa — Console review, May 30, 2024. Simple API to issue search queries to LLMs from code: exa.search(“query here”). Can auto-suggest prompt changes to optimize searches e.g. keyword searches don’t work well with LLMs. Avoids the need for prompt chaining. Has specialist indexes for things like companies, GitHub repos, papers, blogs, jobs.

#### Tradeoffs at review

Exa — Console review, May 30, 2024. SDKs limited to Python, Go, Node.

### Go Micro

Consider for: Agent runtimes and service orchestration in Go

Integration: Go interfaces; MCP and A2A gateways

Reviewed: 2026-06-25

Review: https://console.dev/tools/go-micro#review_1006c6f6e2df9304689388312390f284e9ecb5c9901e9e75b4e9662f3b1eeca1

#### What we liked

Go Micro — Console review, Jun 25, 2026. Supports a variety of runtime primitives including model routing, memory, tools, guardrails, and durable workflows. Automatically creates MCP gateway. Built-in A2A gateway. All abstractions are Go interfaces to make it easily pluggable. Orchestrates multiple services with a local dev server and deploy tooling.

#### Tradeoffs at review

Go Micro — Console review, Jun 25, 2026. Encourages microservices, which are often overkill - especially at the beginning of a new project.

### Langbase

Consider for: Composing LLM workflows and managed retrieval

Integration: Composable pipes; multiple LLM APIs and managed RAG

Reviewed: 2024-08-08

Review: https://console.dev/tools/langbase#review_3d231cb5ecc0551bd891aea24e76aff8bd8787c4c81612e06f6891c9d0320835

#### What we liked

Langbase — Console review, Aug 8, 2024. Composable pipes allow constructing workflows across LLMs. Can be tied into a 10 million token context window memory and a managed RAG store. Supports multiple LLM APIs and provides cost estimates across them so you can easily mix and match depending on budget. JSON mode enforces response format. Usage analytics & logs make it easy to figure out what’s going on.

#### Tradeoffs at review

Langbase — Console review, Aug 8, 2024. Uploaded documents must have selectable text otherwise they can’t be analyzed. Memory only supports plain text formats for now.

### Liveblocks AI Copilots

Consider for: Embedding AI interactions in a React app

Integration: React components; custom tools and your own LLM backend

Reviewed: 2025-06-19

Review: https://console.dev/tools/liveblocks-ai-copilots#review_af306e62d8030cb787207b90f95876c099aa1d15241f676b51722d98dc2212f4

#### What we liked

Liveblocks AI Copilots — Console review, Jun 19, 2025. Drop-in React components for adding AI features to your app. Supports chat, editor toolbar, cursors for streaming editing, and AI submitted comments. Embed app specific knowledge and execute custom tools. Bring your own LLM backend.

#### Tradeoffs at review

Liveblocks AI Copilots — Console review, Jun 19, 2025. Currently closed access - requires you to book a demo to get started.

### Mu

Consider for: Ready-made tools for agents

Integration: MCP server, browser UI and CLI; self-hosted or cloud

Reviewed: 2026-08-06

Review: https://console.dev/tools/mu#review_9763dbcf2998584c4ff9ac18027093e95bb982c3911e11af315f838b63371de8

#### What we liked

Mu — Console review, Aug 6, 2026. MCP server with various tools for agents - news, weather, search, mail. Provides auth and a web UI for using the tools via a browser. Also provided as a CLI. Open source and self-hostable, or available via the author’s cloud service.

#### Tradeoffs at review

Mu — Console review, Aug 6, 2026. Publicly available by default, but has options to turn off signups or just put it on a private network.

### qmd

Consider for: Searching local Markdown from an agent

Integration: On-device CLI and MCP server

Reviewed: 2026-01-22

Review: https://console.dev/tools/qmd#review_d762ccd839607b3324757453e45200891d8d20e6d9bbd5a50376ae6c4628b674

#### What we liked

qmd — Console review, Jan 22, 2026. On-device text search. Indexes anything Markdown. Supports keyword and natural language search. Embeds a local LLM to help with ranking results. Includes an MCP server so you can integrate with your AI of choice. Various output formats (text, JSON, CSV, Markdown).

#### Tradeoffs at review

qmd — Console review, Jan 22, 2026. Not multi-platform - only works on macOS.

### Stagehand

Consider for: Browser automation that combines code and AI

Integration: Playwright-compatible commands with AI fallbacks

Reviewed: 2025-01-16

Review: https://console.dev/tools/stagehand#review_a030d6868bd8c4b5856ed88333f0e6824183c00ea719d975400ddc14ba7db0f3

#### What we liked

Stagehand — Console review, Jan 16, 2025. Makes it easy to build Playwright-compatible browser automations in code. Using AI makes instructions more robust to UI changes. You can provide specific Playwright commands, but use Stagehand as a fallback if they fail. Prompts to extract data can be used to provide useful output.

#### Tradeoffs at review

Stagehand — Console review, Jan 16, 2025. Recommends using Node.js vs newer runtimes like Bun and Deno.

## Decide for your first release

Choose the missing capability in your agent workflow first. These tools solve different problems and are not interchangeable alternatives. The original reviews describe what was available at the time; verify current support before adopting a tool.

### What does your agent need to do?

For retrieval, distinguish web search with Exa from local Markdown search with qmd and a knowledge base with Atomic. For browser actions, start with the Stagehand review. Mu provides ready-made tools, while Go Micro and Langbase address runtime or workflow composition. Liveblocks AI Copilots focuses on the interface inside your app.

### Where will data and actions run?

Decide whether you need local or self-hosted components or a managed service. Check document formats, model-provider access, deployment requirements, and how authentication limits access to tools. Test browser actions and external side effects with explicit permission boundaries.

### What must you verify before shipping?

Try a representative task with your actual stack and data. Check current SDK and platform support, access requirements, costs, and failure recovery. Historical limitations below are dated observations, not claims about the current release.