Voker
Also known as: Voker, Voker.ai, Invoke Labs
Agent analytics platform whose drop-in SDK records every agent call and automatically annotates user intents, corrections and resolutions, reporting resolution and correction rates by agent version.
Voker is an agent analytics platform for teams running AI agents in production. A drop-in Python or TypeScript SDK swaps in for the OpenAI, Anthropic, Gemini or Vercel AI SDK client, or a REST API covers other stacks, and every model call becomes an event tied to an agent version and an end user, with its messages, system prompt, response and tool calls, grouped into sessions.
Voker automatically annotates each event with what the user was trying to do, where they had to correct the agent, and whether the intent was resolved, then groups intents and corrections into categories and reports resolution rates, correction rates and token cost by agent and agent version, so teams can see how a prompt or model change affected each kind of request. Session, agent and person summaries cut down on reading logs. Smart Skills, in beta, generate short prompt amendments from corrections and errors, flag skills that raise cost or correction rates, and version the agent each time one is added or removed.
Voker is aimed at product, engineering, data and leadership teams with production agents handling high conversation volume. Plans run from free at 2,000 events a month to $400 a month for two million events, all with unlimited seats, and an Enterprise plan adds self-hosted deployment, SSO and custom retention.
Vendor details
Canonical URL
https://voker.ai
Category
Agent infrastructure
Subcategory
Agent analytics and observability for AI product teams
Funding status
Independent, legally Invoke Labs, Inc., based in Los Angeles and part of Y Combinator's summer 2024 batch. Disclosed two point two million dollars in pre seed funding from Y Combinator and FundersClub in May 2026, with Dalton Caldwell of Standard Capital and Partner Emeritus at Y Combinator among backers. Co founded by CEO Tyler Postle and CTO Alex Rudolph, who met at a high growth ecommerce startup they helped bootstrap to one hundred million dollars in revenue. Customers include Dutch.com, Lightfield, and True Classic.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
A lightweight, provider agnostic SDK instruments any agent across frameworks, with dedicated wrappers for the leading model providers. It automatically captures events including user messages, assistant messages, and tool calls, tracking both the tools available to an agent and the tools it actually used, and correlates conversational data with the user data teams already collect.
In practice
A hotel booking agent starts getting questions about the attached cafe and restaurant. Voker surfaces that pattern across conversations, so the product team knows to add menu and reservation info instead of guessing.
A recent prompt change quietly breaks an agent's date handling. Voker's abnormality and correction detection catches it in the data before users flood Slack with complaints.
An executive asks how much revenue the agent drives. Voker correlates conversation data with the user data the team already collects, linking agent performance to conversion and retention, so the PM can answer without an engineering ticket.
Sources & related URLs
Related / legacy domains
Agentic Index coverage score
6.0 / 14 capabilities · 43%
| Integrations & Tool Calling | Partial |
|---|---|
|
The SDKs wrap the OpenAI, Anthropic, Gemini and Vercel AI SDK clients and work with LangChain and CrewAI to capture a customer's agent calls and tool calls. A people API ties events to user data the customer already holds. These bring the customer's agent data into Voker. The agent calls its own tools, and Voker names no integration through which an agent reads or writes a real system. SourceVoker, docs.voker.ai overview, provider guides and people conceptread 2026-09-21 |
|
| Workflow Orchestration | Not documented |
|
Customers build and run the agents Voker measures in their own code or in frameworks such as LangChain and CrewAI. Voker does not sequence, branch, retry or route agent steps. SourceVoker, docs.voker.ai overview and voker.ai homepageread 2026-09-21 |
|
| Knowledge Grounding & RAG | Not documented |
|
From their conversations, Voker identifies knowledge gaps in a customer's agents. It names no document ingestion, index, retrieval layer or knowledge API that grounds an agent in company data. SourceVoker, voker.ai homepage and docs.voker.ai overviewread 2026-09-21 |
|
| Human Oversight & Guardrails | Not documented |
|
Agents are measured after the fact, through annotations, rates and session summaries. Human reviewer settings decide whether Voker removes a Smart Skill without review. Voker names no approval step, checkpoint, guardrail or pause on the actions of the customer's agent. SourceVoker, docs.voker.ai Smart Skills and agent analyticsread 2026-09-21 |
|
| Security, Identity & Governance | Partial |
|
SSO comes with the Enterprise plan. Customers retain full ownership of their data, with custom retention on Enterprise. No security attestation is published. SourceVoker, voker.ai homepage and pricingread 2026-09-21 |
|
| Observability & Auditability | Full |
|
Every LLM call or wrapped execution becomes an event tied to an agent version and a person. Voker stores the ordered messages, including user inputs, system prompts and model responses, along with the agent's tool calls. Events group into sessions for a whole conversation or workflow, and teams drill from the dashboard into individual sessions to debug them. Data is kept 30 days on the free plan, 90 days on Starter, one year on Agent First and a custom period on Enterprise. SourceVoker, docs.voker.ai events and sessions, agent analytics, and voker.ai pricingread 2026-09-21 |
|
| Memory & State Persistence | Not documented |
|
For analytics, Voker stores events, sessions and person profiles, and it tracks long running conversations as sessions. It names no session, workflow or long term memory that a customer's agent reads and writes through Voker. SourceVoker, docs.voker.ai events and sessions and long-running conversations cookbookread 2026-09-21 |
|
| Deployment & Data Residency | Full |
|
The platform is hosted, and self hosting is also available. The Enterprise plan includes self hosted deployment and custom data retention. SourceVoker, voker.ai homepage and pricingread 2026-09-21 |
|
| Prebuilt Agents, Templates & Packs | Not documented |
|
Cookbooks show how to instrument customer support and long running agents. Voker names no ready made agents, templates or packaged workflows a customer adopts for its own work. SourceVoker, docs.voker.ai cookbook overviewread 2026-09-21 |
|
| Triggers & Channel Coverage | Full |
|
Annotation runs automatically on every event Voker receives, tagging intents, corrections and resolutions. Voker also generates a Smart Skill for an agent on its own once it reaches a confidence threshold that the prompt is not handling a set of tasks, based on user corrections and logged errors. SourceVoker, docs.voker.ai auto annotations and Smart Skillsread 2026-09-21 |
|
| Model Flexibility & Routing | Not documented |
|
Each agent version's model is recorded, and versions can be compared. The customer's code chooses and calls the model, and Voker offers no model selection, routing, fallback policy or provider configuration of its own. SourceVoker, docs.voker.ai overview and agents conceptread 2026-09-21 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Python and TypeScript SDKs wrap the OpenAI, Anthropic, Gemini and Vercel AI SDK clients. For languages and frameworks the SDKs do not cover, a REST API with OpenAPI specs creates events, people, agents, agent versions and fingerprints. SourceVoker, docs.voker.ai Python SDK, TypeScript SDK and API referenceread 2026-09-21 |
|
| Testing, Debugging & Optimization | Full |
|
Each event is labeled with the user's intent, any correction and whether the intent was resolved. Voker reports resolution and correction rates per intent category and filters them by agent version, so teams compare how a prompt or model change affected each intent. Smart Skills generate prompt amendments from corrections and errors, flag skills that raise cost or correction rate, and version the agent each time. Teams compare those rates across versions to track quality over time and after each deployment. SourceVoker, docs.voker.ai categorization, auto annotations and Smart Skillsread 2026-09-21 |
|
| Browser & Computer Use | Not documented |
|
The product covers analytics, annotations, Smart Skills and SDK instrumentation for AI agents, and Voker names no browser, desktop or computer control by an agent. SourceVoker, docs.voker.ai overview and llms.txtread 2026-09-21 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Free · paid from $80/month
Monthly plans by event allowance, where an event is a user message, assistant message or tool call (about 15 per conversation on average); unlimited seats on every plan; Enterprise custom
What is public
The staged plan structure and a free tier are public, with the Product Hunt listing citing up to two thousand events per month free. Events are defined as user messages, assistant messages, and tool calls. Exact paid rates were not fully retrievable.
Billing mechanics
Self serve staged plans from a free tier upward, billed on events. Start free and scale as agents grow, with team invites. The company targets agents already in production handling at least one thousand conversations per month.
Cost watchouts
Every user message, assistant message and tool call counts as an event, about 15 per conversation, so a busy or tool-heavy agent moves up plans faster than conversation counts suggest; self-hosting and SSO need Enterprise.
Variable cost rationale
Cost steps up by plan as event volume grows, and events count every message and tool call, so spend tracks agent conversation and tool-call volume.
Additional watchouts
Because billing is event based and an event includes each tool call, agents with heavy tool use will accrue events faster than a per conversation view would imply.
Sales call required
Mixed (some tiers require a call)
Free / trial
Free plan at $0 a month with 2,000 events and 30 days of retention
Lowest paid plan
Starter, $80 a month for up to 20,000 events and 90 days of retention
Commercial notes
Self-serve plans from free to $400 a month with unlimited seats, and an Enterprise plan through sales for self-hosting, SSO and custom retention.
Key ambiguities
Overage handling above a plan's event allowance is not stated on the pricing section read.
Missing data
Exact paid tier prices and per event or event allowance details above the free tier were not retrievable from fetched pages.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…
- AlphaBitCore — AI control plane for regulated financial firms: one gateway enforces…
Alternatives to Voker
The closest documented capability profiles to Voker among agent infrastructure platforms tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Hamming AI6.5 / 14Adds documented Human Oversight & Guardrails
- AgentOps5.0 / 14Adds documented Model Flexibility & RoutingVoker vs AgentOps →
- Freeplay8.0 / 14Adds documented Human Oversight & Guardrails and Model Flexibility & Routing
- Monte Carlo8.0 / 14Adds documented Human Oversight & Guardrails and Prebuilt Agents, Templates & Packs
- Traceloop8.0 / 14Adds documented Human Oversight & Guardrails and Model Flexibility & Routing
- Confident AI8.5 / 14Adds documented Human Oversight & Guardrails and Prebuilt Agents, Templates & Packs, among others
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded