Bento Labs
Bento Labs builds self-learning production infrastructure for AI agents, ingesting execution traces to detect drift and continuously update agent behavior. The Y Combinator P26 startup targets engineering teams running agent fleets in production.
- Company typePrivate
- Founded2025
- HeadquartersSan Francisco, United States
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Bento Labs does
Bento Labs (BentoLabs AI, Inc.) is a pre-seed-stage infrastructure company building self-learning production infrastructure for AI agents. Founded in 2025 and headquartered at 360 Baywood Ave, San Jose, CA, the company targets engineering teams running AI agents in production who need continuous learning, observability, and behavioral stability across agent fleets. The platform ingests OpenTelemetry-native traces from agent executions, detects behavioral drift, and runs a recursive learning engine that updates agent policies, prompts, and structured outputs ("Artifacts") automatically over time.
The technical stack centers on three components: a recursive learning engine that closes the loop between observed agent behavior and updated agent logic; "The Book," an evolving knowledge base derived from accumulated traces; and behavioral drift detection that flags regressions in agent performance. The product is monetized via a freemium tier plus usage-based pricing tied to trace and agent-run volume, a structure that supports bottom-up developer adoption with upside into larger commercial contracts. The company is pursuing SOC 2 Type II certification and reports GDPR compliance, positioning for enterprise procurement.
Bento Labs was founded by Abhinav Soni and Kaushik ASP, both previously at Emergent (a Y Combinator S24 company), and was accepted into Y Combinator's P26 (2026) batch. It has raised $500K from Y Combinator and Founder Factor (April 2026) and operates with 1–10 employees. The company shares branding with the BentoML open-source framework, which carries 30K+ GitHub stars and serves as a developer-facing distribution channel; the formal relationship between Bento Labs and the open-source project should be verified.
Bento Labs firmographics
Firmographics- Name
- Bento Labs
- Legal name
- Bento Labs AI, Inc.
- Website
- https://bentolabs.ai
- Company type
- Private
- Founded year
- 2025
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Bento Labs builds self-learning production infrastructure for AI agents, ingesting execution traces to detect drift and continuously update agent behavior. The Y Combinator P26 startup targets engineering teams running agent fleets in production.
- Ownership category
- akta.pro rank
Bento Labs industry classification
Industry- Product category
- AI Agent Observability Platform
- NAICS
- Software Publishers (5132)
- SIC
- Services-Prepackaged Software (7372)
- akta.pro primary industry
- Agents & Autonomous Workflows (Tool Use, Planning, Multi-Agent) (HDAAACAF)
- akta.pro secondary industries
- Model Monitoring, Drift & Behavioral Anomaly Detection (HDAAAKAF), Model Deployment, Serving & Inference Platforms (HDAAABAF), Enterprise Foundation Model Integration & APIs (Connectors, Governance, Deployment) (HDAAACAO)
Keywords
Where Bento Labs is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Bento Labs business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Infrastructure, Marketing or Sales, Operations
Revenue model
- Platform Subscription with Usage-Based Components: Usage-based pricing with flat platform fee. Customers pay for traces ingested and skills promoted, not seats. Free tier for individual developers, startup tier for small teams, and enterprise pricing scaling with trace volume.
- Enterprise Contracts: Enterprise pricing scales with trace volume and includes dedicated support. Multi-year contracts available with volume discounts.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Monthly | Free tier for individual developers |
| Usage-based | Monthly | Startup tier for small teams |
| Usage-based | Annual | Enterprise pricing with dedicated support |
Go-to-market motion2 records
Distribution channels2 records
Marketing channels4 records
Bento Labs product offering
Product offeringCore offering
Bento Labs builds a self-learning production infrastructure platform for AI agents. It captures OpenTelemetry-native traces from every agent framework, detects silent regressions and behavioral drift in real time, and uses a recursive learning engine to convert production failures into durable, reusable skills and knowledge. The platform combines monitoring, alerting, evaluation, and versioned change tracking in a closed-loop system that turns production runs into compounding agent intelligence.
Product overview
Bento Labs offers a unified platform for self-learning production infrastructure for AI agents. The core Bento Platform integrates multiple components including Regression Signals (for training failure detection on production traces), Traces with OpenTelemetry-native span capture, natural language Alerts, and Behavioral Drift detection. The improvement layer comprises Artifacts (reusable skills, subagents, and tools graduated by evidence), The Book (plain-language memory of agent learnings), Evaluations (scoring across offline/CI/live traffic), and Versions (versioned change history). The platform is delivered via an OTel SDK for instrumentation and supports self-hosted deployment options for regulated environments.
Differentiator
Problem solved
Functional benefit
Products and services
- Bento Platform Self-learning production infrastructure for AI agents. Monitors every production run via OpenTelemetry-native traces, detects silent regressions and behavioral drift, and converts production failures into reusable skills, subagents, and tools through a recursive learning engine. Includes natural-language alerts, evaluations, versioned change tracking, and a plain-language agent memory (The Book).
- Bento OTel SDK OpenTelemetry-native Python SDK that instruments AI agents built on LangChain, LlamaIndex, Vercel AI SDK, AutoGen, custom frameworks, or raw provider calls. Captures spans, tool calls, tokens, and outcomes with full fidelity for the Bento platform.
- Self-Hosted Deployment Fully self-hosted deployment of the Bento platform running the collector, storage, and recursive learning engine inside a customer's VPC or on-prem environment, designed for regulated industries and enterprise security requirements.
Quantifiable outcome
- 2.6× higher scores on ARC-AGI-3 with self-learning layer, same agent same budget, 34% cheaper per successful outcome
- +3 more outcomes
Companies that use Bento Labs
Customer profileSegments3 records
Ideal customer profiles3 records
Bento Labs technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration11 records
AI capability11 records
Feature11 records
Bento Labs partnerships and signals
Strategic signalScale indicators3 records
Recent moves6 records
Expansion highlights5 records
Bento Labs competitors and assessment
Company assessmentDirect peers
- Helicone: LLM observability and monitoring platform offering tracing, evaluation, and usage analytics. Targets the same enterprise teams running AI agents in production that Bento serves.
- LangSmith: LangChain's native observability and evaluation platform for LLM/agent applications. Directly comparable to Bento's trace, eval, and monitoring surfaces — Bento explicitly positions as a complement that closes the feedback loop LangSmith leaves open.
- Arize AI: Provides LLM and ML observability, evaluation, and drift detection. Overlaps directly with Bento's behavioral drift detection, regression signals, and trace-based monitoring for production AI systems.
- Langfuse: Open-source LLM/agent observability and evaluation platform. Closely comparable to Bento's core tracing and evaluation features; Bento differentiates via its closed-loop recursive learning engine on top of similar trace data.
- WhyLabs: AI observability platform with drift detection, model monitoring, and data quality features. Comparable to Bento's behavioral drift detection and regression signals for production AI systems.
Broad incumbents
- Honeycomb: Production observability platform with high-cardinality tracing. Overlaps with Bento's OTel-native trace capture and span-tree debugging but is not agent-specialized.
- Datadog: Broad cloud observability platform expanding into LLM/AI monitoring (Datadog LLM Observability). Competes with Bento at the observability layer but lacks an agent-specific closed-loop learning engine.
Emerging players
- Confident AI (DeepEval): Open-source LLM evaluation framework (DeepEval) with managed cloud offering. Comparable to Bento's Evaluations module but lighter on production monitoring and recursive learning.
- Braintrust: Evaluation and observability platform purpose-built for LLM applications. Overlaps with Bento's Evaluations and Traces features; less differentiated on closed-loop learning and behavioral drift.
- Maxim AI: Evaluation, simulation, and observability platform for AI agents. Comparable in evaluation and monitoring functionality, but with a stronger focus on pre-release simulation rather than Bento's production closed-loop learning.
Market position
Strengths5 records
Weaknesses4 records
Competitive moat5 records
Key risks5 records
Key highlights6 records
Customer concentration
Bento Labs social profiles
Digital presenceBento Labs compliance and trust
Trust signalCompliance2 records
Bento Labs financial estimates
Financial estimateRevenue estimate
Valuation estimate
Bento Labs leadership team
Management profileNumber of profiles
Profiles2 records
Bento Labs funding detail
Funding detailFunding overview
Funding rounds1 record
Investors2 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Bento Labs M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Bento Labs
What does Bento Labs do?
Bento Labs builds a self-learning production infrastructure platform for AI agents. It captures OpenTelemetry-native traces from every agent framework, detects silent regressions and behavioral drift in real time, and uses a recursive learning engine to convert production failures into durable, reusable skills and knowledge. The platform combines monitoring, alerting, evaluation, and versioned change tracking in a closed-loop system that turns production runs into compounding agent intelligence.
Is Bento Labs a public or private company?
Bento Labs is a private company. It is classified as venture growth investor backed and is currently operating.
When was Bento Labs founded?
Bento Labs was founded in 2025. It employs 11 to 50 people.
Where is Bento Labs based?
Bento Labs is headquartered in San Francisco, United States, in the North America region.
How does Bento Labs make money?
Two revenue lines are on record. Platform Subscription with Usage-Based Components are the primary driver. The others are enterprise Contracts.
Who are Bento Labs's main competitors?
Direct peers on record are Helicone, LangSmith, Arize AI, Langfuse and WhyLabs. Broad incumbents are Honeycomb and Datadog. Emerging players are Confident AI (DeepEval), Braintrust and Maxim AI.
Does Bento Labs have an API?
Yes. Bento Labs offers an OpenTelemetry (OTel) SDK for instrumenting AI agents. The SDK enables full-fidelity trace capture across agent frameworks. The company also mentions AI Tools that use artificial intelligence models and supports third-party AI Tools that customers can select or enable.
What industry is Bento Labs in?
Bento Labs's product category is AI Agent Observability Platform. Its primary akta.pro industry code is HDAAACAF, Agents & Autonomous Workflows (Tool Use, Planning, Multi-Agent), with a secondary code of HDAAAKAF, Model Monitoring, Drift & Behavioral Anomaly Detection. Its NAICS code is 5132 and its SIC code is 7372.