AgentOps
Session replay and cost tracking for AI agent debugging
Overview
AgentOps is a developer SDK for observing AI agents — it traces every LLM and tool call, tracks token cost across 400+ models, and replays a run step by step. Integrates with CrewAI, LangChain, AutoGen, the OpenAI Agents SDK and more; free up to 5,000 events, though the SDK has not had a release since August 2025.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Agent-agnostic SDK
- LLM cost tracking (400+ models)
- Replay analytics
- Unlimited events and log retention
- Session/event export
- Role-based permissions
- Dedicated Slack + email support
- SLA, custom SSO, Slack Connect
- On-premise deployment and custom retention
- SOC-2, HIPAA, NIST AI RMF
Use cases
What you can produce with AgentOps
- Session replay / time-travel debugging
- Multi-agent workflow visualisation
- Token and LLM-cost tracking across 400+ models
- SDK integrations: CrewAI, AutoGen, LangChain, OpenAI Agents SDK, Agno, AG2, CamelAI, Google ADK
- MIT-licensed SDK, with a self-hostable backend under Elastic License 2.0
- MCP server and read-only public API
Once an agent makes more than one LLM call, you lose visibility fast — which step looped, which tool call blew the budget, why the output drifted. AgentOps answers that with session replay, per-call cost and token tracking and multi-agent visualisation, wired in with a few lines of SDK code across an unusually wide set of frameworks. One independent benchmark measured about 12% runtime overhead, worth knowing before you instrument a latency-sensitive path. The concern now is maintenance: the SDK has not shipped a release since August 2025 and commits have slowed to a trickle, in a category where frameworks change monthly and better-funded rivals such as Langfuse and LangSmith keep shipping. The free tier covers 5,000 events and Pro starts at $40 a month. Try it if its framework coverage fits your stack, but check the repository's recent activity before you build on it.
What we found
AgentOps is an observability SDK for AI agents: a few lines of code record every LLM call, tool call and error in a run, track token cost across more than 400 models, and let you replay the session step by step — the vendor calls it time-travel debugging. It plugs into a wide spread of frameworks, including CrewAI, Agno, the OpenAI Agents SDK, LangChain, AutoGen, AG2 and CamelAI, with newer documented integrations for Google's ADK, LangGraph, Smolagents, Mem0 and xAI, plus an MCP server and a read-only public API. That breadth is the usual reason third-party roundups recommend it for teams running more than one agent framework.
Performance evidence is limited. AIMultiple's benchmark of agent-monitoring tools, which ran 100 identical queries through a travel-planning workflow, measured about 12% runtime overhead for AgentOps against 15% for Langfuse — a single test, but the only independent number we found.
The more important 2026 finding is momentum. The SDK's last release, version 0.4.21, shipped on 29 August 2025; the repository has seen only three commits since October 2025, the latest in June 2026, and community pull requests opened in late September 2026 sit unmerged. Nothing indicates a shutdown: the repository is not archived, the homepage still sells the product and the dashboard still accepts sign-ins. But for a tool that must track fast-moving agent frameworks, a year without a release is a real risk to weigh.
The company, which trades as Agency, was co-founded by Alex Reibman, Adam Silverman and Shawn Qiu in San Francisco and raised a $2.6 million pre-seed round led by 645 Ventures and Afore Capital in 2024. The SDK is MIT-licensed and the backend is published under the Elastic License 2.0, with a public self-hosting guide. Pricing: Basic is free for up to 5,000 events; Pro starts at $40 a month pay-as-you-go with unlimited events, unlimited retention, export and role-based permissions; Enterprise adds SLAs, SSO, on-premise deployment and SOC 2, HIPAA and NIST AI RMF support.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including AgentOps's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with AgentOps unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to AgentOps directly →
Spotted something out of date? Suggest an update →
Alternatives to AgentOps
More in Agents & Automation
