Top 9 of 33 alternatives
9 LiveKit Agents alternatives (2026)
The closest alternatives to LiveKit Agents in October 2026 are ElevenLabs Agents, TelEcho and AutoGen.
9 agents compared · 6 with a free plan · prices checked 2 September–24 October 2026
Reviewed by The Agents Index editorial team.
LiveKit Agents and its alternatives at a glance
| Agent | Best for | Starts at | Better than LiveKit Agents at | Checked |
|---|---|---|---|---|
| LiveKit Agents | Engineering teams that want to own their voice-agent stack end to end, self-hosted or on LiveKit Cloud, without being locked into one vendor's API. | Free, then $50/month source | 2 September 2026 | |
| ElevenLabs Agents | Developers and product teams who want ElevenLabs' voice quality and cloning as the foundation of their agent, and are comfortable assembling the rest of the stack themselves. | Free, then $6/month source | Done-for-you service, Agent permissions | 18 October 2026 |
| AutoGen | Python or .NET developers and researchers exploring multi-agent orchestration who want a mature, free, self-hosted framework. | Free source | Bring your own key, Getting out | 19 September 2026 |
| Retell AI | Product and engineering teams that want a low-latency, human-sounding phone agent with flexible model and telephony choice. | From $0.07/minute source | Done-for-you service, MCP server | 23 September 2026 |
| OpenAI Agents SDK (Python) | Developers who want the fastest, lightest way to ship a straightforward agent (support triage, a tool-using assistant or a sandboxed coding agent) with tracing and guardrails included. | Free source | Agent permissions, Bring your own key | 18 October 2026 |
| CrewAI | Developers who want to stand up a role-based, collaborating multi-agent system quickly in Python, with a stable MIT-licensed core and a hosted build and runtime for production. | Free source | Done-for-you service, MCP server | 13 September 2026 |
| Letta | Developers and researchers who specifically need an agent whose memory, skills and identity genuinely persist and evolve across long-running sessions, rather than reset on every call. | Free, then $20/month source | Agent permissions, In your pipeline | 2 September 2026 |
| Microsoft Agent Framework | .NET or Python teams already invested in Azure or Microsoft Foundry who want one supported agent SDK, an optional managed hosting path, and a documented migration off AutoGen or Semantic Kernel. | Free source | Agent permissions, Bring your own key | 19 September 2026 |
| OpenAI Realtime API | Engineering teams building a voice agent from scratch who want to go directly to OpenAI's own speech-to-speech model (no orchestration platform, no per-minute markup) and are comfortable being single-vendor on OpenAI. | From $10/1m audio input tokens source | Agent permissions, Pooled budget | 23 October 2026 |
Shortlist
Closest alternatives to LiveKit Agents
The ones people open most here come first, then the closest matches. How we rank
- 01
ElevenLabs Agents
Hosted platform for building voice and chat agents, with telephony, SDKs and a choice of LLM.
From FreeWhy it's on this listDone-for-you service · PartlySelf-serve software you run, but enterprises can buy ElevenLabs Forward Deployed Engineers who scope, build and stabilise the deployment; no price is published.Source ↗
Where it improves on LiveKit Agents
- +Agent permissions · YesTool use is gated per MCP server and per tool: always ask, fine-grained (auto-approved, requires approval, disabled), or no approval at all.Source ↗
- +In your pipeline · YesAgents deploy from CI with an API key in the environment; the unit of change is the agent configuration, not a pull request in your code.Source ↗
Trade-offs against LiveKit Agents
- −Self-hosted · PartlyThe full agent runtime can run in your own VPC or hardware on AWS or GCP, but the documentation is NDA-gated and sold through the account team; Azure is unreleased.Source ↗
- −Open source · PartlyThe CLI and the client SDKs are public on GitHub and take contributions; the agent platform and the speech models themselves are proprietary and hosted.Source ↗
- 02
TelEcho
Conversational AI platform for voice and messaging agents that run customer calls and update connected systems.
From $0.07/minuteWhy it's on this listDone-for-you service · PartlyYou run the software yourself in a no-code builder, but the vendor also sells hands-on help: strategy calls, an onboarding meeting and prompt-engineering support appear only in the Enterprise column of the comparison table.Source ↗
Where it improves on LiveKit Agents
- +Agent permissions · YesScope is bounded at setup rather than per action: guardrails, escalation and handoff conditions, business hours and response policies are configured before an agent goes live, and no per-step approval prompt is described.Source ↗
- +MCP server · YesTelEcho documents a native MCP integration letting an agent inside ChatGPT or Claude define and configure TelEcho agents; no server address, tool inventory or auth flow is published to verify scope.Source ↗
Trade-offs against LiveKit Agents
- −Claude · PartlyClaude appears as a place you drive TelEcho from, via the vendor's MCP integration, not as a model the agent can run on; the pricing calculator's LLM selector offers OpenAI models only.Source ↗
- −Open models · NoNo open-weight option is offered anywhere: the calculator's five LLM choices are all OpenAI, and the vendor's complete sub-processor inventory names OpenAI as the only LLM provider that touches customer data.Source ↗
- −Model choice · PartlyYou choose the telephony carrier, the STT engine, the TTS voice and the execution pipeline, but every LLM option in the calculator comes from one vendor, so model choice stops at OpenAI.Source ↗
- 03
AutoGen
Open-source Microsoft framework for building single and multi-agent AI applications in Python.
From FreeWhy it's on this listBring your own key · YesYou bring your own inference throughout: your OpenAI key, your Azure deployment and AAD identity, or a local Ollama server on your own hardware.Source ↗
Where it improves on LiveKit Agents
- +Getting out · YesTeams, agents and conditions serialise to JSON via dump_component and save_state, and Studio persists sessions in your own SQLite or PostgreSQL database. No credits exist to expire.Source ↗
Trade-offs against LiveKit Agents
- −Claude · PartlyA first-party Anthropic client exists in the extensions package, but Microsoft labels the section experimental rather than a supported integration.Source ↗
- −Open models · PartlyOpen-weight models run through the Ollama client or OpenAI-compatible endpoints, but that client is marked experimental and Microsoft warns local models are less capable.Source ↗
- −On the command line · PartlyA built-in Console UI streams a run to the terminal and UserProxyAgent can take console input, but there is no standalone agent CLI; the autogenstudio command only launches the web UI.Source ↗
- 04
Retell AI
Pay-as-you-go platform for building, testing and running AI phone, chat and SMS agents.
From $0.07/minuteWhy it's on this listDone-for-you service · PartlyYou run the software yourself in Retell's no-code studio; hands-on implementation by Retell's people is sold only as an Enterprise add-on, not as a managed service.Source ↗
Where it improves on LiveKit Agents
- +MCP server · YesRetell runs a hosted MCP endpoint authenticated with an API key, exposing agent, call, knowledge-base and QA tools so another assistant can drive the account.Source ↗
- +Pooled budget · YesA workspace holds one prepaid credit balance that every agent and member draws down in real time, and it blocks all new calls when it empties.Source ↗
Trade-offs against LiveKit Agents
- −Open models · PartlyRetell's own model menu is limited to hosted GPT, Claude and Gemini, but the custom-LLM path lets you put a self-hosted open-weights model behind the agent over a WebSocket you operate.Source ↗
- −On the command line · NoThe docs enumerate every way to drive Retell: dashboard, REST API, Node/Python SDKs and MCP server. No command-line tool is among them.Source ↗
- −Self-hosted · NoThe platform runs on Retell's own infrastructure: pay-as-you-go uses shared servers and the most isolation offered is a dedicated Retell server sold as an Enterprise add-on, with no customer-infrastructure deployment documented.Source ↗
- 05
OpenAI Agents SDK (Python)
Open-source Python framework for building agents with tools, handoffs, guardrails and tracing.
From FreeWhy it's on this listAgent permissions · YesTools declare when they need approval, the run pauses and surfaces interruptions, and the serialized run state can be approved or rejected before resuming, including for shell, apply_patch and MCP tools.Source ↗
Where it improves on LiveKit Agents
- +Bring your own key · YesYou always bring your own inference credentials, and any OpenAI-compatible endpoint can be pointed at by setting base URL and key, so the vendor never resells inference through this package.Source ↗
Trade-offs against LiveKit Agents
- −Claude · PartlyThe SDK names only OpenAI models; non-OpenAI providers are reached through OpenAI-compatible endpoints or the beta Any-LLM and LiteLLM adapters, and Claude is never named in the docs.Source ↗
- −Open models · PartlyAny OpenAI-compatible endpoint or the beta Any-LLM/LiteLLM adapters can be used, which covers self-hosted open-weight serving, but no open-weight model or local runtime is named in the docs.Source ↗
- −In a browser · PartlyThere is no hosted web application for building or running agents; the only web surface is OpenAI's Traces dashboard, where runs recorded by built-in tracing can be inspected.Source ↗
- 06
CrewAI
Open-source multi-agent framework with a managed platform for deploying and governing agent workflows.
From FreeWhy it's on this listDone-for-you service · PartlyYou run the software yourself, but the Enterprise plan sells forward-deployed engineers, onboarding and training separately, so people-delivered build help exists as a paid extra.Source ↗
Where it improves on LiveKit Agents
- +MCP server · YesBoth plans can publish a built workflow as an MCP server so other agents can call it, and agents can equally consume any external MCP server as tooling.Source ↗
- +Bring your own key · YesInference runs on your own provider accounts: you supply each provider key, or manage LLM connections centrally in the platform, and Enterprise can host the runtime in your cloud.Source ↗
Trade-offs against LiveKit Agents
- −How it meters · PartlyMetering is per workflow execution rather than per seat, and the free tier states its allowance, but Enterprise volumes are sized case by case with unpriced overage.Source ↗
- −Opt out of training · PartlyFramework telemetry excludes prompts and agent data by default and is disabled with one environment variable, but no page states whether platform execution data trains vendor models.Source ↗
- 07
Letta
Stateful agent harness, cloud and SDK for agents with persistent, git-versioned memory.
From FreeWhy it's on this listAgent permissions · YesFour modes from strict to unrestricted, plus allow/deny tool patterns and a cross-agent memory guard.Source ↗
Where it improves on LiveKit Agents
Trade-offs against LiveKit Agents
- −MCP server · NoMCP runs inbound only: Letta sessions consume stdio, HTTP and SSE MCP servers; no Letta MCP server is published.Source ↗
- −Pooled budget · NoQuota attaches to each purchased Pro seat; unpaid seats get no Letta-hosted inference at all.Source ↗
- −Audit log · NoNo administrative audit log is published; the Enterprise list stops at RBAC, SSO, quotas and dedicated support.Source ↗
- 08
Microsoft Agent Framework
Open-source .NET, Python and Go SDK for building AI agents and graph-based multi-agent workflows.
From FreeWhy it's on this listAgent permissions · YesApproval is per tool and configurable in code: the local shell tool asks for every command by default, regex allow and deny lists pre-filter commands, and disabling approval requires an explicit unsafe acknowledgement.Source ↗
Where it improves on LiveKit Agents
- +Bring your own key · YesYou bring your own inference: the framework requires you to configure your chosen provider's keys and endpoints, and it never brokers or resells model access.Source ↗
- +Getting out · YesNothing is locked in a vendor store: sessions and history go to a session store you implement over your own database or blob storage, and there are no credits to expire.Source ↗
Trade-offs against LiveKit Agents
- −On the command line · PartlyA terminal surface exists only as developer tooling: the DevUI command line launches agents from a directory, and the Harness console apps are explicitly samples rather than shipped components.Source ↗
- −In a browser · PartlyDevUI gives a browser interface for running and debugging agents, but the vendor labels it a sample app unfit for production; shipping a real web surface means adding the preview AG-UI or ChatKit integrations.Source ↗
- −How it meters · PartlyThe framework itself costs nothing, so all spend is model and tool consumption metered by whichever provider you configure; Microsoft publishes no calculator and states the usage cost is yours.Source ↗
- 09
OpenAI Realtime API
Speech-to-speech API for voice agents, with tools, MCP and WebRTC or WebSocket transport.
From $10/1m audio input tokensWhy it's on this listAgent permissions · YesPermissioning is explicit and developer-held: function tools exist precisely so approval checks and private system access stay in your application, while MCP tools let the API call a remote server for you.Source ↗
Where it improves on LiveKit Agents
- +Pooled budget · YesSpend pools at the organisation, with an approved usage limit covering it, and projects underneath can carry their own custom rate and spend limits as sub-pools.Source ↗
- +Admin controls · YesAdministrators get real levers rather than advice: enforced maximum API key lifetimes, key-type restrictions down to disabling new keys, hard monthly spend caps, and per-project rate and spend limits.Source ↗
Trade-offs against LiveKit Agents
- −Runs autonomously · PartlyThe session handles audio turns, interruptions, tool calls and handoffs without a human prompting each step, but the vendor's own guidance keeps approval checks, guardrails and blocking in the developer's application.Source ↗
- −Claude · NoOnly OpenAI's own realtime snapshots are supported; the vendor's complete per-endpoint model inventory for /v1/realtime names six gpt-realtime models and no third-party model.Source ↗
- −Open models · NoNo open-weight option on this endpoint: the vendor's exhaustive supported-snapshot list for /v1/realtime contains only its own hosted gpt-realtime models.Source ↗
What to watch for
- LiveKit AgentsA framework, not a no-code or dashboard-first builder. Expect to write and deploy real code, unlike this category's more configured rivals.
- ElevenLabs AgentsCovers the voice + LLM-orchestration layer, not a turnkey ops product, CRM/helpdesk workflows, observability and compliance UI are yours to build.
- AutoGenIt is a code library, not a hosted service. You build, host and operate the agents yourself.
- Retell AITrue per-minute cost is the sum of stacked components, so budgeting requires modelling model, voice, and telephony together.
- OpenAI Agents SDK (Python)No automatic step-by-step checkpointing, `RunState` only snapshots at human-in-the-loop approval pauses, so you manage durable state yourself outside of that.
- CrewAIPricing jumps directly from a free 50-execution Basic tier to custom Enterprise, with no self-serve plan in between
- LettaSmaller ecosystem and community than LangChain, LlamaIndex or Microsoft AutoGen.
- Microsoft Agent FrameworkIt is a code library plus an optional managed layer, not a no-code product. You (or Foundry Hosted Agents) run and operate it.
- OpenAI Realtime API60-minute maximum session duration.
Side by side
Compare LiveKit Agents with the shortlist
Compare alternatives to LiveKit Agents
The digest
Which agents actually ship. New agents and articles, by email, at most once a week.
Questions buyers ask
What are the best alternatives to LiveKit Agents?
ElevenLabs Agents, TelEcho and AutoGen are the closest. They share the most of LiveKit Agents's features.
What is the cheapest alternative to LiveKit Agents?
ElevenLabs Agents, AutoGen and OpenAI Agents SDK (Python) and 3 more have a free plan.
Which agents here have a free plan?
6 of the 9: ElevenLabs Agents, AutoGen, OpenAI Agents SDK (Python), CrewAI and Letta and 1 more.
When were these prices checked?
Between 2 September 2026 and 24 October 2026, each read on the maker's own pricing page.
Sources and method
Closest first: most of LiveKit Agents's features, then the fewest gaps. How we rank
Every price here was read on the maker's own pricing page.
- LiveKit Agents: livekit.com/pricing, checked 24 October 2026
- ElevenLabs Agents: elevenlabs.io/pricing/agents, checked 2 September 2026
- TelEcho: telecho.io/pricing, checked 16 September 2026
- AutoGen: github.com/microsoft/autogen/blob/main/LICENSE-CODE, checked 19 September 2026
- Retell AI: retellai.com/pricing, checked 23 September 2026
- OpenAI Agents SDK (Python): raw.githubusercontent.com/openai/openai-agents-python/main/LICENSE, checked 20 September 2026
- CrewAI: crewai.com/pricing, checked 13 September 2026
- Letta: docs.letta.com/pricing, checked 24 October 2026
- Microsoft Agent Framework: github.com/microsoft/agent-framework/blob/main/LICENSE, checked 19 September 2026
- OpenAI Realtime API: developers.openai.com/api/docs/pricing, checked 23 October 2026
The digestFree
Which agents actually ship.
New agents and articles, by email, at most once a week.
Last issue · No. 1 · 13 October 2026Devin vs Claude Code: Which Should You Pick?
One-click unsubscribe
The Agents Index digest
The agents we added and what we wrote about them, by email, no more than once a week.
Last issue: Devin vs Claude Code: Which Should You Pick? · 13 October 2026