Ops and automation
Make
Make is a visual automation tool. You drag modules into a scenario that moves data between apps, including AI steps.
- workflows
- integrations
Freemiummake.com
How to build a business agent system
Start from one workflow with a measurable outcome, such as resolved tickets or hours saved. Customer support, sales ops, back office and finance, and an internal or IT helpdesk are the jobs this guide is aimed at. Anthropic's engineering post says: "the most successful implementations weren't using complex frameworks." Start simple, and add autonomy only when you need it.
The groups below are the stack: no-code workflows, agent SDKs, a durable runtime, memory, evals, a gateway for guardrails and spend, a support agent you can buy, and connections to other tools. Tools are alphabetical inside each group. This is not a scored ranking.
Building effective agents · Anthropic
How to choose
Use n8n, Zapier, or Make when the workflow should be a no-code scenario with an AI step. Use LangGraph, CrewAI, the OpenAI Agents SDK, or the Claude Agent SDK when the agent is code. Use Temporal or Inngest when a run has to retry, resume, or stay on a schedule. Use Mem0 or Zep when the agent has to recall facts, or keep your own store. Use Langfuse, Braintrust, or Arize AX when you need a trace and an eval. Use Portkey when you need budgets, fallbacks, caching, or guardrails in front of the model. Use Intercom Fin when the job is resolving support tickets and handing the rest to a person. Use Bivia when the agent needs a typed decision and a confidence score, including over MCP.
The same jobs and layers are listed as use-case categories. Sales ops, back office, and IT helpdesk do not have a verified tool in this directory yet.
Workflow automation with AI steps
n8n, Zapier, and Make are the no-code tools named for workflows with AI steps. n8n can be self-hosted, and its pricing includes a free Community Edition. Zapier's description includes Zapier Agents, Chatbots, and Zapier MCP. Make is a visual scenario builder billed in credits. Paid prices for Make were not visible on the pricing page at the last check, so the listing quotes the free plan only.
Ops and automation
n8n
n8n connects apps and AI steps into workflows you build on a canvas. You can self-host it or use n8n Cloud, under a public license that is not standard open source.
- workflows
- self-hosted
- integrations
Freemiumn8n.ioOps and automation
Zapier
Zapier connects apps so a trigger in one can run actions in another. Its AI tools help you build those workflows and run agents.
- workflows
- integrations
- agents
Freemiumzapier.com
Orchestration and agent SDKs
LangGraph, CrewAI, and the OpenAI Agents SDK were already listed. The Claude Agent SDK is Claude Code's agent loop as a Python and TypeScript library, with tools, hooks, subagents, MCP, permissions, and sessions. CrewAI also has a hosted platform. The OpenAI Agents SDK and CrewAI describe human-in-the-loop checks, and the Claude Agent SDK describes permissions.
Agent frameworks
Claude Agent SDK
The Claude Agent SDK is Claude Code's agent loop as a Python and TypeScript library, with tools, hooks, subagents, MCP, permissions, and sessions.
- agents
- Python
- MCP
Freedocs.claude.comAgent frameworks
Open sourceCrewAI
CrewAI lets you define agents with roles and tasks that they work through together. The framework is open source, with a paid cloud for running crews.
- multi-agent
- Python
Open sourcecrewai.comAgent frameworks
Open sourceLangGraph
LangGraph is a library for agents that follow a graph of steps, with pauses and retries. It comes from the LangChain team.
- agents
- workflows
- Python
Open sourcelangchain.comAgent frameworks
Open sourceOpenAI Agents SDK
The OpenAI Agents SDK is a Python library for agents that call tools, hand work to other agents, and check results. It is open source.
- Python
- OpenAI
- tools
Open sourceopenai.github.io
Durable runtime
Temporal and Inngest run work durably. Temporal can run OpenAI Agents SDK and Google ADK model and tool calls as durable activities, with an open-source self-hosted option. Inngest is an open-source SDK for event-driven functions and agents, with retries, concurrency, tracing, and sandboxes.
Agent frameworks
Open sourceInngest
Inngest is an open-source SDK for event-driven durable functions and agents, with retries, concurrency, tracing, and sandboxes.
- agents
- durable execution
Freemiuminngest.comAgent frameworks
Open sourceTemporal
Temporal is durable workflow execution, and it can run OpenAI Agents SDK and Google ADK model and tool calls as durable activities.
- workflows
- durable execution
Freemiumtemporal.io
Memory and context
Mem0 is a memory layer for agents, with an MCP integration, OpenMemory, and graph memory on custom plans. Zep is a temporal context graph for agent memory, with a Context MCP Server. Letta was already listed for agents that keep memory over time. The same layer can be your own store, for example Notion or a database.
Agent frameworks
Open sourceLetta
Letta, formerly MemGPT, is an open source platform for agents that keep memory over time. You can run it yourself or use the hosted service.
- memory
- agents
- self-hosted
Open sourceletta.comAgent frameworks
Mem0
Mem0 is a memory layer for agents, with an MCP integration, OpenMemory, and graph memory on custom plans.
- memory
- MCP
Freemiummem0.aiAgent frameworks
Zep
Zep is a temporal context graph for agent memory, and it includes a Context MCP Server for that memory.
- memory
- context
Freemiumgetzep.com
Evals and observability
Langfuse is open-source tracing, prompt management, evals, and cost tracking, and self-hosting is free. Braintrust is a tracing and evaluation platform with scorers, human review, and a Loop agent. Arize AX traces agents and runs evals with OpenTelemetry, and Phoenix is open source and local-first. LangSmith tracing, evals, deployment, and an LLM gateway are part of the LangChain platform already listed as LangChain and LangGraph.
Agent frameworks
Open sourceArize AX / Phoenix
Arize AX does agent tracing and evals with OpenTelemetry, and Phoenix is the open-source local-first product.
- tracing
- evals
Freemiumarize.comAgent frameworks
Braintrust
Braintrust is a tracing and evaluation platform with scorers, human review, and a Loop agent.
- evals
- tracing
Freemiumbraintrust.devAgent frameworks
Open sourceLangfuse
Langfuse is open-source tracing, prompt management, evals, and cost tracking, and you can self-host it for free.
- tracing
- evals
Freemiumlangfuse.com
Guardrails, approvals, and cost control
Portkey is an AI gateway with fallbacks, caching, budgets, guardrails, observability, and an MCP gateway, and an open-source gateway is available. n8n describes human approval for tool calls. Those workflow and SDK listings stay in the groups above.
A support agent you can buy
Intercom Fin resolves support tickets across chat, email, and phone, and it works with existing helpdesks. Its pricing page lists $0.99 per Fin outcome. Customer figures quoted by Intercom are labelled as vendor claims later on this page.
MCP servers and tool connectivity
Composio connects an agent to apps such as GitHub, Slack, and Gmail and handles sign-in. Pomerado turns a website into an API and MCP server an agent can call. Bivia returns a typed decision and a confidence score, and the product page says you can add it as an MCP server in a coding agent. Zapier MCP is part of the Zapier description in the workflow group.
Agent frameworks
Bivia
Bivia is Articul8's multimodal model that returns a typed decision and a confidence score, and it can abstain when the evidence is not enough. You can call it over HTTP or add it as an MCP server in a coding agent.
- decisions
- MCP
- API
Paidarticul8.aiAgent frameworks
Open sourceComposio
Composio connects an agent to apps such as GitHub, Slack, and Gmail, and handles sign-in for those tools. The SDK is open source, and the hosted service has a free tier.
- tools
- integrations
- auth
Freemiumcomposio.devAgent frameworks
Open sourcePomerado
Pomerado turns a website into an API and MCP server an agent can call, including sites that need a login. The core is open source under the MIT license, and the hosted service can sign in, including 2FA, and repair an integration when the site changes.
- MCP server
- websites
- agents
Freemiumpomerado.ai
Production patterns
- Put an approval in front of an irreversible action, including an outbound message or a spend.
- Keep a trace for every run.
- Build eval datasets from real failures.
- Notice when the price is per outcome. Intercom lists Fin at $0.99 per outcome.
- Retries need idempotency, so a step that runs twice does not do the work twice.
- Give the agent least-privilege credentials for the tools it can call.
Vendor claims
These figures are vendor claims. AgentAtlas has not measured them.
Robin, an Intercom customer: "50% resolution rate".
Vendor claim, quoted by Intercom. AgentAtlas has not measured this.
Synthesia, an Intercom customer: resolution time down 96%.
Vendor claim, quoted by Intercom. AgentAtlas has not measured this.
Buy or build
Hosted plans are in the buy column. Libraries and self-host options are in the build column. Prices are the ones on each listing.
| Layer | Buy | Build or self-host |
|---|---|---|
| Workflow | Zapier, Make, or n8n Cloud. | n8n's free Community Edition. The pricing page says a standard version can be self-hosted. |
| Agent logic | CrewAI's hosted platform, which lists a free plan of 50 workflow executions a month and custom Enterprise pricing. | LangGraph, the CrewAI framework, the OpenAI Agents SDK, or the Claude Agent SDK. |
| Runtime | Temporal Cloud, or Inngest's hosted plans. | Temporal's open-source server, or Inngest's open-source SDK. |
| Memory | Mem0 or Zep hosted plans. | Your own store, for example Notion or a database. |
| Evals | Langfuse's hosted plans, Braintrust, or Arize AX. | Self-hosted Langfuse, which is free, or Phoenix, which is open source. |
| Cost control | Portkey Production, listed at $49 per month. The Developer plan is free and includes 10,000 logs. | Portkey's open-source gateway. |
Categories
Other guides
- Best AI workflow automation
- Best frameworks for building agents
- Best open-source AI tools to run yourself
20 tools in this guide on AgentAtlas. Each card links to the listing, and the listing links to the official site.