Category · Agent frameworks
Agent frameworks and SDKs
Libraries that run the agent loop. Tool calling, MCP, multi-agent hand-offs, durable state, human approval and tracing. Compared on what they do by default, including the telemetry they send.
Best agent frameworks and SDKs, the shortlist171 head-to-head comparisons
Capability keys agent.framework · agent.multi-agent · agent.durable · agent.mcp-client · All tools
letme.dev/agent.framework picks the top-graded tool in this list and says how to call it direct (how letme works).
The same listing from the live API. Graded results come first, then the official MCP registry when no graded-only filter is set.
https://www.anchorterminal.com/api/v1/search
Filters
| Compare | # | Tool | Category | Grade | Score | Agent rating | p95 | Context | Price / x402 | Auth | Where | Details |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | AI SDKVercel · Agent framework | Frameworks | AA | 93.6 | none | n/a | n/a | Free · OSS | API key | Library | ||
|
Open-source TypeScript toolkit from Vercel for building AI applications and agents, with a provider-agnostic API, a ToolLoopAgent, an MCP client and UI hooks for React, Vue, Svelte and Angular. Top strength Apache 2.0, with 246 npm releases of Top weakness TypeScript only, and |
||||||||||||
| 3 | OpenAI Agents SDKOpenAI · Agent framework | Frameworks | AA | 90 | 3.9 (8) | n/a | n/a | Free · OSS | API key | Library | ||
|
Multi-agent framework built on agents, hand-offs and guardrails, with sessions, tracing and human approval. Top strength MCP in about 11 lines, with static and dynamic tool filters and require_approval Top weakness Tracing on by default, with model and tool content, sent to OpenAI |
||||||||||||
| 7 | Pydantic AIPydantic · Agent framework | Frameworks | AA | 85.6 | 4.0 (8) | n/a | n/a | Free · OSS | None | Library | ||
|
Typed Python agent framework for 25+ model providers, with MCP, A2A and durable execution. Top strength Typed outputs and tools, validated by Pydantic, with failed validations sent back to the model Top weakness Python only, and 3.11 or newer since 2.55.0 |
||||||||||||
| 12 | Strands AgentsAmazon Web Services · Agent framework | Frameworks | A | 83.7 | none | n/a | n/a | Free · OSS | None | Library | ||
|
Strands Agents is AWS's open-source agent SDK for Python and TypeScript. A model-driven loop runs tools, MCP servers, multi-agent patterns, sessions, interrupts for human approval and OpenTelemetry tracing, with Amazon Bedrock as the default model provider. Top strength No usage analytics found, with OpenTelemetry tracing opt-in Top weakness Three advisories in October 2026 on A2A session sharing and world-readable default session files, fixed in 1.43.0 and 1.48.0 |
||||||||||||
| 15 | Microsoft Agent FrameworkMicrosoft · Agent framework | Frameworks | A | 83.1 | 3.0 (4) | n/a | n/a | Free · OSS | None | Library | ||
|
Microsoft Agent Framework is an open-source SDK for building AI agents and multi-agent workflows in Python and .NET, with a Go version in public preview. Microsoft names it the successor to AutoGen and Semantic Kernel. Top strength MIT, with Python 1.24.0 and .NET 1.27.0 both tagged on 30 October 2026, past 1.0 since 2 April 2026 Top weakness Changes marked [BREAKING] shipped in minor releases of released packages, including agent-framework-core in 1.22.0 on 15 October and 1.23.0 on 22 October 2026 |
||||||||||||
| 32 | AgentOSFramers Lab, Inc. · Agent framework | Frameworks | A | 79.5 | 2.8 (4) | n/a | n/a | Free · OSS | API key | Library | ||
|
Open-source TypeScript agent framework from Framers Lab, installed from npm as @framers/agentos. It runs agents with long-term memory, multi-agent teams, graph workflows with checkpoints, guardrails and approval gates across 11 model providers. Top strength Apache 2.0, with no analytics or telemetry code found in the source and a privacy page that says the runtime sends nothing to the vendor Top weakness Version 0.15.69, with 95 release tags after 0.14.2 to 1 November 2026 and breaking minors 0.14.0 and 0.15.0 in that time |
||||||||||||
| 59 | MastraKepler Software, Inc. · Agent framework | Frameworks | A | 78 | none | n/a | n/a | Free · OSS | API key | Library | ||
|
Open-source TypeScript framework from Mastra for building agents and workflows, with memory, RAG, evals, tracing, an MCP client and server, and a local Studio. An optional hosted platform runs deployments and observability. Top strength MCP client with requireToolApproval per server or per tool, allowedHosts and a curated environment for stdio servers Top weakness Malicious versions of Mastra npm packages, including @mastra/core 1.42.1, were published on 17 June 2026 |
||||||||||||
| 68 | Docker AgentDocker · Agent framework | Frameworks | BB | 77.4 | 2.8 (4) | n/a | n/a | Free · OSS | None | Library | ||
|
Docker's open-source runtime for building and running AI agents from YAML or HCL files, formerly cagent. It runs as a `docker agent` CLI plugin with a terminal UI, headless mode, HTTP, MCP and A2A servers, and a Go library. Top strength An agent with one MCP server is eight lines of YAML, with a published JSON Schema for the config Top weakness Usage telemetry is on by default and command events can include prompts passed as arguments |
||||||||||||
| 157 | AgnoAgno Inc. · Agent framework | Frameworks | BB | 74 | 3.3 (4) | n/a | n/a | Free · OSS | API key | Library | ||
|
Open-source Python framework from Agno Inc. for building agents, teams and workflows, with the AgentOS runtime that serves them over a REST API and an MCP server. Formerly Phidata. Top strength Apache 2.0, with 27 stable releases tagged in the 90 days to 1 November 2026 and 3.1.12 on 29 October Top weakness Usage telemetry is on by default and sends session, run and agent identifiers unhashed. Prompts and outputs are not sent |
||||||||||||
| 165 | LlamaIndex FrameworkLlamaIndex · Agent framework | Frameworks | BB | 73.8 | none | n/a | n/a | Free · OSS | None | Library | ||
|
The LlamaIndex Framework is LlamaIndex's open-source Python toolkit for retrieval and agents, with FunctionAgent, ReActAgent and CodeActAgent, multi-agent AgentWorkflow, event-driven Workflows, MCP tools and hundreds of integration packages. Top strength FunctionAgent, ReActAgent, CodeActAgent and multi-agent AgentWorkflow over one Workflows engine Top weakness Core is 0.14.x, and minor versions such as 0.13 and 0.14 carried breaking changes |
||||||||||||
| 171 | LangGraphLangChain · Agent framework | Frameworks | BB | 73.6 | 3.5 (2) | n/a | n/a | Free · OSS | None | Library | ||
|
Low-level runtime for long-running, stateful agent graphs with checkpointed persistence, human-in-the-loop, memory and streaming. Top strength Checkpointed persistence, so runs survive restarts and resume where they stopped Top weakness No MCP client of its own. LangChain's langchain.mcp is in beta |
||||||||||||
| 204 | Claude Agent SDKAnthropic · Agent framework | Frameworks | BB | 72.8 | 2.5 (4) | n/a | n/a | Free | API key | Library | ||
|
Runs the Claude Code agent as a library, with its loop, built-in tools, permissions, sessions, hooks and subagents. Top strength Claude Code's file, shell, search and web tools, with six permission modes, a canUseTool callback and PreToolUse hooks Top weakness Claude models only, so a person has to set up a Claude API key or cloud account first |
||||||||||||
| 217 | Agent Development Kit (ADK)Google · Agent framework | Frameworks | BB | 72.4 | 3.0 (8) | n/a | n/a | Free · OSS | None | Library | ||
|
Google's code-first toolkit to build, evaluate and deploy agents in Python, TypeScript, Go, Java and Kotlin. Top strength Python, TypeScript, Go, Java and Kotlin, with Python 2.13.0 on 23 October and Java 1.13.0 on 21 October 2026 Top weakness Six CVEs in 2026, three of them critical, one in tool confirmation itself |
||||||||||||
| 353 | BoxLang AIOrtus Solutions · Agent framework | Frameworks | B | 69 | none | n/a | n/a | Free · OSS | None | Library | ||
|
BoxLang AI (bx-ai) is Ortus Solutions' Apache-2.0 agent framework for BoxLang on the JVM, with agents, tools, memory, RAG, human approval, and an MCP client and server across more than 15 model providers. Top strength Human approval enforced in code, with suspended runs that resume without replaying the model call, five approval policies and durable grants Top weakness Runs only on BoxLang, a JVM language, so it suits teams already on BoxLang or CFML rather than Python or TypeScript stacks |
||||||||||||
| 519 | Cursor Agent SDKCursor · Agent framework | Frameworks | B | 65.5 | none | n/a | n/a | Your plan | OAuth or key | Library | ||
|
Cursor's TypeScript framework for defining an agent as a folder of instructions, typed tools, skills, MCP connections and channels, testing it with evals, and running it on your own server or on Cursor's hosting. Top strength Tools declare a read, write or idempotent effect. Only reads and idempotent calls retry, and dry-run sessions stub writes Top weakness Closed source under Cursor's terms, with no public repository, CI or issue tracker |
||||||||||||
| 537 | TradingAgentsTauric Research · Agent framework | Frameworks | B | 65.1 | none | n/a | n/a | Free · OSS | API key | Library | ||
|
TradingAgents, from Tauric Research, is an open-source Python framework on LangGraph in which LLM agents for analysis, research debate, trading and risk review produce a five-tier rating and a report for one ticker and date. Top strength Apache 2.0, with seven releases between 31 August and 27 October 2026 and "Upgrading from" notes in a dated changelog Top weakness No MCP client or MCP server, so an agent drives it only through Python or the CLI |
||||||||||||
| 667 | smolagentsHugging Face · Agent framework | Frameworks | B | 62.5 | none | n/a | n/a | Free · OSS | None | Library | ||
|
smolagents is Hugging Face's small Python library for agents that write and run Python code (CodeAgent) or call tools as JSON (ToolCallingAgent), with MCP tools, any model provider and sandboxed execution through E2B, Docker, Modal or Blaxel. Top strength CodeAgent and ToolCallingAgent in a small, readable codebase Top weakness No release since 1.26.0 on 29 May 2026, and few commits to main since August |
||||||||||||
| 712 | CrewAICrewAI · Agent framework | Frameworks | C | 61.5 | 3.0 (2) | n/a | n/a | Free · OSS | None | Library | ||
|
Python framework for teams of role-playing agents in Crews, plus event-driven Flows for stateful workflows. Top strength An MCP server in five lines through the mcps field, with static and dynamic tool filters Top weakness Anonymous telemetry on by default, with no stated destination or retention |
||||||||||||
| 831 | NitroStackNitrostack Inc. · Agent framework | Frameworks | C | 58.6 | 3.8 (4) | n/a | n/a | Freemium | None | Library | ||
|
NitroStack is a TypeScript framework for MCP servers, with schema validation, auth modules and tool annotations. NitroStudio supports inspection and testing, while NitroCloud hosts deployed servers. Top strength Apache-2.0 TypeScript source with Zod input and output schemas for tools Top weakness The CLI sends PostHog usage events by default with no opt-out, described only in a source-tree file |
||||||||||||
| 885 | guilds.runCigale AI · Agent framework | Frameworks | C | 57.2 | none | n/a | n/a | Free · OSS | API key | Library | ||
|
Open-source, self-hosted system from Cigale AI for standing teams of AI agents, each with its own role, model and tools. A guild shares chat, notes and tables, and every run keeps the prompt, tool calls and memory changes. Top strength Every run stores the prompt the model saw, each tool call and result, and versioned memory with the author of each change Top weakness No tagged release, version number or published image on 4 November 2026. Users install whatever is on main, including breaking changes |
||||||||||||
Nothing matches these filters. .
p95 latency and context cost come from our probes, which haven't run yet, so those columns start hidden. Grades run from AA to F, and agent-ready means BB or better. Filters, sorting and export run in your browser; the table is complete without JavaScript.
Indexed, not reviewed (1)
Listings sorted into this category from public catalogues (the official MCP registry, APIs.guru, the x402 Bazaar and OpenRouter), with facts and our own checks but no score, grade or rank. How the index works.
| Listing | Kind | What it does | Why it's here |
|---|---|---|---|
| FlurryPORT flurryport.io | MCP server | Webhook capture and replay, delivery pipes, and signed multi-agent rooms. No signup to start. | vendor's own |
How the ranking works
Every tool is scored 0 to 100 across 9 weighted categories, minus deductions for negative events, and given a grade from AA to F. Reliability, latency and context cost are measured by probes and by traffic through letme. Schema quality, security and payments are assessed against a published checklist. Task success comes from a fixed task suite. Payments, including x402, are one category of the nine. Read the methodology.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.