Activepieces
Build automations with app integrations, AI steps, and reusable pieces.
Explore it forConnect business applications and add AI to repeatable workflows.
Setup & source
Cloud workspace or self-hosted instance, plus credentials for connected apps.
Format: Visual workflow builder.
Check which integrations and administration features are included in your chosen edition.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
Arize Phoenix
Inspect agent traces, evaluate outputs, and compare changes using datasets and experiments.
Explore it forDebug tool calls and retrieval, then test whether a change improves results.
Setup & source
A Phoenix instance and instrumentation that sends application traces.
Format: Tracing & evaluation platform.
Phoenix and Arize AX are separate offerings; choose the deployment and data controls you need.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
Atlassian MCP Server
Project tracking · Vendor-maintained
Search Jira and Confluence, and create or update work items and pages.
Explore it forTurn project context and meeting notes into tracked work.
Access, compatibility & source
- Maintainer
- Atlassian (vendor-maintained)
- Connection
- Remote HTTP
- Compatible clients
- Claude, Cursor, VS Code, and other documented MCP clients
- Authentication
- OAuth 2.1 or API token, subject to admin policy
- Permissions & actions
- Existing user access and admin-enabled permission groups; write and delete tools depend on grants.
- Getting started
- Atlassian Cloud account and a supported MCP client
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
Braintrust
Collect agent traces, review failures, and run evaluations to compare application changes.
Explore it forTurn production examples into repeatable tests for prompts and agents.
Setup & source
Braintrust account and an instrumented application or evaluation dataset.
Format: Observability & evaluation platform.
Choose representative datasets and scorers; review plan limits and trace data handling.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
Claude Code
Work through coding tasks with an agent in your terminal or editor.
Explore it forExplore a repository, implement changes, and iterate on code.
Setup & source
Account and development environment.
Format: Terminal & IDE. Check the provider for current pricing, limits, and permissions.
Official source ↗Source checked 18 Oct 2026. Not hands-on tested.
Cline
Read and edit files, run commands, and use browser tools from a coding agent.
Explore it forWork through repository changes with configurable model access and approval controls.
Setup & source
Install a supported Cline application and configure model access.
Format: Editor, terminal & desktop.
Model access can use Cline billing or your own provider credentials; review execution permissions.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
Cloudflare MCP Server
Infrastructure · Vendor-maintained
Discover and call Cloudflare API operations from an AI assistant.
Explore it forInspect infrastructure and automate account operations, including Workers workflows.
Access, compatibility & source
- Maintainer
- Cloudflare (vendor-maintained)
- Connection
- Remote HTTP
- Compatible clients
- HTTP MCP clients supporting OAuth or bearer tokens
- Authentication
- OAuth or a scoped Cloudflare API token
- Permissions & actions
- Access follows selected OAuth permissions or API-token grants; can modify infrastructure when authorized.
- Getting started
- Cloudflare account and an HTTP MCP client
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
Codex
OpenAI’s coding agent for working with codebase context, editing files, and running development tools.
Explore it forImplement features, investigate bugs, and review code in local or cloud workflows.
Setup & source
Sign in and select a project or configure a cloud environment.
Format: Desktop, CLI, IDE & cloud. Check the provider for current pricing, limits, and permissions.
Official source ↗ · Setup options ↗Source checked 18 Oct 2026. Not hands-on tested.
Context7 MCP
Documentation · Vendor-maintained
Retrieve library documentation and code examples relevant to your task.
Explore it forCheck version-specific APIs while implementing a feature.
Access, compatibility & source
- Maintainer
- Upstash (vendor-maintained)
- Connection
- Remote HTTP or local stdio
- Compatible clients
- Cursor, Claude Code, VS Code, and other documented MCP clients
- Authentication
- API key or OAuth; access and quotas depend on configuration
- Permissions & actions
- Documentation lookup; private repositories require account authorization. Does not itself edit your code.
- Getting started
- Configure a remote endpoint or the local Node.js package
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
CrewAI
Code framework · Multi-agent collaboration
Organize agents into collaborative crews and explicit workflows.
Explore it forCoordinate specialist agents around a repeatable business process.
Capabilities, deployment & source
- Type
- Code framework
- Languages / scope
- Python
- Model support
- Multiple providers through configured LLM integrations.
- State, recovery & approvals
- Flows, state, checkpointing, and human-review steps; configure recovery for your workflow.
- Tools & integrations
- Function tools, knowledge sources, MCP, and A2A support.
- Testing & deployment
- Run in your Python infrastructure; the separate AMP platform adds managed operations.
- License
- MIT core; commercial platform terms are separate.
- Consider before choosing
- Crew and Flow abstractions require deliberate task boundaries and coordination.
- Getting started
- Python environment, model credentials, and tool configuration.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
Cursor
An AI coding environment for working with your codebase and delegating development tasks.
Explore it forWrite and revise code with repository context.
Setup & source
Install and sign in.
Format: Editor & CLI. Check the provider for current pricing, limits, and permissions.
Official source ↗Source checked 18 Oct 2026. Not hands-on tested.
Figma MCP Server
Design · Vendor-maintained
Bring design context into coding workflows and use supported tools to work with Figma files.
Explore it forTranslate a design into implementation context or connect designs to code.
Access, compatibility & source
- Maintainer
- Figma (vendor-maintained)
- Connection
- Remote HTTP; desktop option available
- Compatible clients
- Approved clients including Claude Code, Cursor, and VS Code
- Authentication
- Figma sign-in and authorization for the remote server
- Permissions & actions
- File access and available actions depend on your account, seat, and tools; remote tools can also write designs.
- Getting started
- Figma account and a client in Figma’s MCP catalog
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
Gemini CLI
Use Gemini from the terminal to understand code, edit projects, and automate tasks.
Explore it forCombine local project context, shell tools, and MCP integrations in a terminal workflow.
Setup & source
Install Gemini CLI and authenticate with Google sign-in, a Gemini API key, or Vertex AI.
Format: Terminal agent.
Authentication options have different access requirements and quotas; review sandbox settings.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
GitHub Copilot
Coding assistance and agent workflows connected to your editor and GitHub.
Explore it forBring AI assistance into an existing GitHub development workflow.
Setup & source
GitHub account and supported environment.
Format: IDE & GitHub. Check the provider for current pricing, limits, and permissions.
Official source ↗Source checked 18 Oct 2026. Not hands-on tested.
GitHub MCP Server
Code & delivery · Vendor-maintained
Connect an assistant to repository content, issues, pull requests, and GitHub workflows.
Explore it forInvestigate an issue or review a pull request with repository context.
Access, compatibility & source
- Maintainer
- GitHub (vendor-maintained)
- Connection
- Remote HTTP or local server
- Compatible clients
- VS Code, Cursor, Claude Code, and other documented MCP clients
- Authentication
- OAuth where supported, or a GitHub personal access token
- Permissions & actions
- Repository access follows credentials and enabled tools; read-only mode excludes write tools.
- Getting started
- MCP client and GitHub authorization
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
Google ADK
Code framework · Agent workflows
Build and evaluate agents with code-defined workflows and tools.
Explore it forCombine agents, deterministic steps, and external tools in an application.
Capabilities, deployment & source
- Type
- Code framework
- Languages / scope
- Python edition covered here; other language SDKs available.
- Model support
- Optimized for Gemini; supports other models through integrations.
- State, recovery & approvals
- Workflow state, loops, retries, and human input; runtime and SDK configuration determine persistence.
- Tools & integrations
- Function tools, OpenAPI, and MCP integrations.
- Testing & deployment
- Local development and evaluation; container deployment or Google Cloud services.
- License
- Apache-2.0 (Python SDK).
- Consider before choosing
- Check feature parity in your chosen language SDK and deployment runtime.
- Getting started
- Install the SDK and configure model access and tools.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
Haystack
Code framework · Retrieval & agent pipelines
Compose retrieval pipelines and tool-calling agents from reusable components.
Explore it forBuild an assistant that searches documents and acts through tools.
Capabilities, deployment & source
- Type
- Code framework
- Languages / scope
- Python
- Model support
- Multiple model providers through generator components and integrations.
- State, recovery & approvals
- Agent state and pipeline breakpoints/snapshots; configure application recovery and approvals.
- Tools & integrations
- Document stores, retrievers, component tools, and MCPToolset.
- Testing & deployment
- Self-managed Python applications; tracing integrations and evaluation components.
- License
- Apache-2.0.
- Consider before choosing
- Component wiring and retrieval design require application-level engineering.
- Getting started
- Python environment, model provider, and retrieval or tool components.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
Kiro
Develop with an agent using specifications, project steering, hooks, and connected tools.
Explore it forTurn requirements into a structured implementation plan and work through its tasks.
Setup & source
Install Kiro and sign in with a supported account.
Format: IDE & terminal.
Choose the IDE or CLI workflow that fits your team; check account limits and tool permissions.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
Langfuse
Trace agent behavior, evaluate outputs, and manage prompts in one platform.
Explore it forUnderstand quality, latency, and cost across AI application traces.
Setup & source
Instrument your application.
Format: Platform & SDK. Check the provider for current pricing, limits, and permissions.
Official source ↗Source checked 18 Oct 2026. Not hands-on tested.
LangGraph
Orchestration runtime · Stateful orchestration
Define stateful agent execution with explicit control over workflow steps.
Explore it forBuild long-running workflows that pause for review and resume from saved state.
Capabilities, deployment & source
- Type
- Orchestration runtime
- Languages / scope
- Python and JavaScript/TypeScript; details here follow Python docs.
- Model support
- Model-agnostic; call your chosen model inside graph nodes.
- State, recovery & approvals
- Checkpointers, durable execution, retry policies, and human-in-the-loop interrupts.
- Tools & integrations
- Custom nodes and tools; model and MCP adapters through the LangChain ecosystem.
- Testing & deployment
- Run in your infrastructure; LangSmith provides separate tracing and deployment services.
- License
- MIT core; hosted services have separate terms.
- Consider before choosing
- Low-level graph control means you design state, persistence, and execution boundaries.
- Getting started
- Application code, model configuration, and a persistence backend where needed.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
Linear MCP Server
Project tracking · Vendor-maintained
Find, create, and update issues, projects, and comments in Linear.
Explore it forTurn a feature plan into issues or summarize delivery progress.
Access, compatibility & source
- Maintainer
- Linear (vendor-maintained)
- Connection
- Remote HTTP
- Compatible clients
- Claude, Cursor, VS Code, and other documented MCP clients
- Authentication
- OAuth 2.1, bearer token, or Linear API key
- Permissions & actions
- Read-write by default; a read-only endpoint or restricted read scope is available.
- Getting started
- Linear workspace and an HTTP MCP client
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
LlamaIndex
Code framework · Knowledge & document agents
Connect agents to document retrieval, data sources, and event-driven workflows.
Explore it forBuild a knowledge assistant using your own documents and retrieval tools.
Capabilities, deployment & source
- Type
- Code framework
- Languages / scope
- Python edition covered here; a TypeScript library is also available.
- Model support
- Multiple providers through model and embedding integrations.
- State, recovery & approvals
- Workflow events, context/state, step retry policies, and human-input patterns.
- Tools & integrations
- Data connectors, indexes, vector stores, and tool integrations.
- Testing & deployment
- Run the library yourself; LlamaParse is a separate document-processing platform.
- License
- MIT framework; hosted document services have separate terms.
- Consider before choosing
- Retrieval quality still depends on parsing, indexing, and evaluation choices.
- Getting started
- Python environment, data sources, and model/embedding configuration.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
Mastra
Code framework · AI application development
Build agents and workflows within a TypeScript application stack.
Explore it forAdd tool-using agents to a web product with workflow and memory support.
Capabilities, deployment & source
- Type
- Code framework
- Languages / scope
- TypeScript
- Model support
- Multiple model providers through its model-routing integrations.
- State, recovery & approvals
- Workflow branching, retries, suspend/resume, and storage-backed execution state.
- Tools & integrations
- Application tools, MCP clients/servers, retrieval, and storage adapters.
- Testing & deployment
- Self-host as an application/server; Studio, evals, tracing, and a separate hosted platform.
- License
- Apache-2.0 core; ee/ features use the Mastra Enterprise License.
- Consider before choosing
- Choose storage and deployment deliberately; enterprise features have separate licensing.
- Getting started
- TypeScript project, model credentials, and storage for persistent workflows.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
n8n
Connect applications and AI steps in visual automation workflows.
Explore it forAutomate repeatable work across services and APIs.
Setup & source
Configure connections and credentials.
Format: Visual workflow builder. Check the provider for current pricing, limits, and permissions.
Official source ↗Source checked 18 Oct 2026. Not hands-on tested.
Notion MCP
Knowledge & collaboration · Vendor-maintained
Search, read, create, and update Notion workspace content.
Explore it forRetrieve team knowledge and turn findings into pages or databases.
Access, compatibility & source
- Maintainer
- Notion (vendor-maintained)
- Connection
- Remote HTTP
- Compatible clients
- Claude Code, Cursor, and other documented MCP clients
- Authentication
- OAuth connection to your Notion workspace
- Permissions & actions
- Can read and update content you can access; workspace controls may restrict connected clients.
- Getting started
- Notion account and a compatible MCP client
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
OpenAI Agents SDK
Code SDK · Tool use & agent handoffs
Build agent loops with tools, handoffs, guardrails, and run tracing.
Explore it forCoordinate specialist agents with explicit tool access and human approval.
Capabilities, deployment & source
- Type
- Code SDK
- Languages / scope
- Python and TypeScript; detailed capabilities vary by SDK.
- Model support
- OpenAI models by default; custom model/provider integrations are supported.
- State, recovery & approvals
- Sessions, turn limits, approval interruptions, and resumable run state; configure durable infrastructure separately.
- Tools & integrations
- Function tools, MCP servers, and supported hosted tools.
- Testing & deployment
- Run SDK code in your infrastructure; built-in tracing and evaluation integration.
- License
- MIT SDK; model, hosted-tool, and runtime usage are separate.
- Consider before choosing
- Hosted capabilities and provider compatibility vary; the SDK is not a managed deployment service.
- Getting started
- Python or TypeScript environment and model-provider credentials.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
OpenCode
Work on code with an open-source agent that supports multiple model providers and MCP tools.
Explore it forUse a terminal-centered coding workflow while keeping model-provider choice.
Setup & source
Install OpenCode and connect a supported model provider.
Format: Terminal, desktop & web.
V1 and V2 have different installation and migration guidance; this listing links to V2 documentation.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
Playwright MCP
Browser automation · Vendor-maintained
Give an assistant browser navigation, interaction, and inspection tools through Playwright.
Explore it forExplore a web application and reproduce a browser workflow.
Access, compatibility & source
- Maintainer
- Microsoft (vendor-maintained)
- Connection
- Local stdio; HTTP option available
- Compatible clients
- VS Code, Cursor, Claude Code, and other documented MCP clients
- Authentication
- No service API key for local use; target sites may require login
- Permissions & actions
- Controls a browser and can act through its logged-in sessions; this is not a read-only connector.
- Getting started
- Node.js, an MCP client, and a browser environment
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
Promptfoo
Evaluate and red-team LLM applications with repeatable tests.
Explore it forCompare prompts and models, and check application behavior in CI.
Setup & source
Test cases and model configuration.
Format: CLI & library. Check the provider for current pricing, limits, and permissions.
Official source ↗Source checked 18 Oct 2026. Not hands-on tested.
Pydantic AI
Code SDK · Typed agent applications
Build typed agents with validated outputs, tools, and configurable model providers.
Explore it forIntegrate structured agent results into a Python application.
Capabilities, deployment & source
- Type
- Code SDK
- Languages / scope
- Python
- Model support
- Multiple providers and custom models through a common typed interface.
- State, recovery & approvals
- Validation retries, deferred tool approvals, persistence capabilities, and optional durable-execution engines.
- Tools & integrations
- Typed function tools, dependency injection, MCP, and durable runtime integrations.
- Testing & deployment
- Self-managed Python deployment; Pydantic Evals and OpenTelemetry/Logfire integration.
- License
- MIT SDK; hosted services have separate terms.
- Consider before choosing
- Durable execution and telemetry need their own backend configuration.
- Getting started
- Python environment, provider access, and typed tools/output schemas.
Official documentation ↗ · Repository & license ↗Sources checked 18 Oct 2026. Documentation reviewed; no hands-on benchmark. Verify features for your SDK version.
Sentry MCP
Observability · Vendor-maintained
Search errors, analyze performance, triage issues, and manage Sentry projects.
Explore it forInvestigate a production error using its issue and performance context.
Access, compatibility & source
- Maintainer
- Sentry (vendor-maintained)
- Connection
- Remote HTTP
- Compatible clients
- Claude Code, Cursor, VS Code, and compatible HTTP MCP clients
- Authentication
- OAuth with your Sentry account
- Permissions & actions
- Uses account access; organization or project scoping narrows the connection. Includes management actions.
- Getting started
- Sentry account and an HTTP MCP client
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
Supabase MCP Server
Databases · Vendor-maintained
Connect an assistant to Supabase project and database tools.
Explore it forExplore database schemas and support application development.
Access, compatibility & source
- Maintainer
- Supabase (vendor-maintained)
- Connection
- Remote HTTP
- Compatible clients
- Claude Code, Cursor, and compatible authenticated HTTP MCP clients
- Authentication
- Browser sign-in granting access to a Supabase organization
- Permissions & actions
- Scope to a project and select tool groups; enable read-only mode when inspecting data.
- Getting started
- Supabase account, project, and a compatible MCP client
Official source ↗Source checked 18 Oct 2026. Documentation reviewed; connection not hands-on tested. Check current plans and client requirements with the provider.
Windmill
Turn scripts into workflows, internal applications, and AI-powered automation.
Explore it forRun scheduled or event-driven engineering workflows with reusable code steps.
Setup & source
Cloud workspace or self-hosted instance, plus scripts and service credentials.
Format: Code-first workflows & apps.
Self-hosting includes operating workers and storage; review edition-specific controls.
Official source ↗Source checked 6 Nov 2026. Documentation reviewed; not hands-on tested. Check current pricing and availability with the provider.
