Kumo.AI

Kumo.AI builds predictive AI for relational data. Its flagship model, KumoRFM, is a pre-trained Relational Foundation Model that generates training-free predictions over multi-table data by interpreting a data warehouse as a temporal heterogeneous graph, queried through a SQL-like Predictive Query Language (PQL). Kumo exposes this through a Python SDK (kumoai), a REST API for automating retraining and batch predictions, an open-source Model Context Protocol server (kumo-rfm-mcp) for agentic workflows, and a published Agent Skills catalog. It ships as SaaS, VPC/BYOC, a Snowflake Native Application, and a Databricks Native Application. Kumo product documentation is now hosted by NVIDIA under Structured Data and Graph Models.

Kumo.AI publishes 2 APIs on the APIs.io network. Tagged areas include Company, Ai, Machine Learning, Predictive Analytics, and Graph Neural Networks.

Kumo.AI’s developer surface includes documentation, API reference, getting-started guide, support, signup flow, authentication, changelog, and 20 more developer resources.

35.6/100 thin Agent 44/100 agent aware Full breakdown ↓
scored 2026-07-20 · rubric v0.4
2 APIs 2 MCP Servers 0 Features
CompanyAiMachine LearningPredictive AnalyticsGraph Neural NetworksFoundation ModelsData WarehouseModel Context ProtocolAgentsRelational Data

API Rating

API Evangelist API Evangelist Rating How this is scored →
scored 2026-07-20 · rubric v0.4
Composite quality — 35.6/100 · thin
Contract Quality 0.0 / 25
Developer Ergonomics 14.3 / 20
Commercial Clarity 7.9 / 20
Operational Transparency 4.1 / 13
Governance 0.0 / 12
Discoverability 9.3 / 10
Agent readiness — 44/100 · agent aware
Machine-Readable Contract 0 / 18
Agentic Access Contract 0 / 15
MCP Server 12 / 12
Machine-Readable Auth 10 / 10
Idempotency 0 / 9
Stable Error Semantics 8 / 8
Request/Response Examples 0 / 7
Rate-Limit Signaling 7 / 7
Typed Event Surface 0 / 6
Agent Skills 5 / 5
Well-Known Catalog 4 / 4
Consent & Bot Identity 0 / 3
Improve this rating by publishing the missing artifacts — every area above can be raised, and the full rubric is at apis.io/rating/. This rating is computed from github.com/api-evangelist/kumoai: open an issue to ask a question, or submit a pull request to add artifacts. Want it done for you? Prioritized profiling — $2,500 →

APIs

Kumo REST API

Automate and schedule predictive query training and batch predictions. Retrain a predictive query (re-ingesting data from the source), trigger recurring batch predictions, monit...

KumoRFM MCP Server

Open-source Model Context Protocol server that empowers AI assistants with KumoRFM. Builds, manages, and visualizes relational graphs directly from CSV or Parquet files, convert...

MCP Servers

kumoai-mcp.yml

MCP SERVER

kumoai-mcp.yml

MCP SERVER

Rate Limits

Kumoai Rate Limits

0 limits

RATE LIMITS

Resources

Get Started 3

Portal, sign-up, and the first successful call

Documentation 2

Reference material describing how the API behaves

Agent Surfaces 4

MCP servers, agent skills, and machine-readable catalogs

Design & Contract 4

Pagination, idempotency, versioning, errors, and events

Build 3

SDKs, sample code, and the tooling you integrate with

Access & Security 6

Authentication, authorization, and security posture

Operate 3

Status, limits, changes, and where to get help

Commercial 1

Pricing, plans, and the legal terms of use

Company 1

The organization behind the API

Source (apis.yml)

apis.yml Raw ↑
aid: kumoai
name: Kumo.AI
description: Kumo.AI builds predictive AI for relational data. Its flagship model, KumoRFM, is a pre-trained Relational Foundation
  Model that generates training-free predictions over multi-table data by interpreting a data warehouse as a temporal heterogeneous
  graph, queried through a SQL-like Predictive Query Language (PQL). Kumo exposes this through a Python SDK (kumoai), a REST
  API for automating retraining and batch predictions, an open-source Model Context Protocol server (kumo-rfm-mcp) for agentic
  workflows, and a published Agent Skills catalog. It ships as SaaS, VPC/BYOC, a Snowflake Native Application, and a Databricks
  Native Application. Kumo product documentation is now hosted by NVIDIA under Structured Data and Graph Models.
url: https://raw.githubusercontent.com/api-evangelist/kumoai/refs/heads/main/apis.yml
image: https://avatars.githubusercontent.com/u/83320328?v=4
x-type: company
x-source: vc-portfolio
x-backed-by:
- sv-angel
x-tier: enriched
x-tier-reason: enrichment-pipeline
specificationVersion: '0.20'
created: '2026-07-17'
modified: '2026-07-19'
tags:
- Company
- Ai
- Machine Learning
- Predictive Analytics
- Graph Neural Networks
- Foundation Models
- Data Warehouse
- Model Context Protocol
- Agents
- Relational Data
apis:
- aid: kumoai-rest-api
  name: Kumo REST API
  description: Automate and schedule predictive query training and batch predictions. Retrain a predictive query (re-ingesting
    data from the source), trigger recurring batch predictions, monitor job status and evaluation metrics, and cancel in-progress
    jobs. Authenticates with a tenant-scoped X-API-Key header.
  humanURL: https://docs.nvidia.com/sdgm/rest-api
  baseURL: https://<customer_id>.kumoai.cloud/api
  x-baseURL-note: The public API URL is tenant-scoped; SPCS deployments use https://<kumo_spcs_deployment_url>/api.
  tags:
  - Machine Learning
  - Batch Prediction
  - Jobs
  properties:
  - type: Documentation
    url: https://docs.nvidia.com/sdgm/rest-api
  - type: Authentication
    url: ../authentication/kumoai-authentication.yml
- aid: kumoai-rfm-mcp
  name: KumoRFM MCP Server
  description: Open-source Model Context Protocol server that empowers AI assistants with KumoRFM. Builds, manages, and visualizes
    relational graphs directly from CSV or Parquet files, converts natural language into PQL queries, and runs training-free
    predictions, evaluations, and explanations. Runs over stdio or streamable-http and is distributed on PyPI and as an MCPB
    bundle.
  humanURL: https://github.com/kumo-ai/kumo-rfm-mcp
  baseURL: https://kumorfm.ai
  tags:
  - Model Context Protocol
  - Agents
  - Predictions
  properties:
  - type: SourceCode
    url: https://github.com/kumo-ai/kumo-rfm-mcp
  - type: MCPServer
    url: ../mcp/kumoai-mcp.yml
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com
- FN: APIs.json
  email: info@apis.io
common:
- type: Website
  url: https://kumo.ai
- type: DeveloperPortal
  url: https://docs.nvidia.com/sdgm/rfm/overview
- type: Documentation
  url: https://docs.nvidia.com/sdgm/rfm/introduction
- type: APIReference
  url: https://docs.nvidia.com/sdgm/rest-api
- type: GettingStarted
  url: https://docs.nvidia.com/sdgm/quick-start/rfm
- type: Support
  url: https://join.slack.com/t/kumoaibuilders/shared_invite/zt-2z9uih3lf-fPM1z2ACZg~oS3ObmiQLKQ
- type: GitHubOrganization
  url: https://github.com/kumo-ai
- type: SignUp
  url: https://kumorfm.ai
- type: PrivacyPolicy
  url: https://docs.nvidia.com/sdgm/privacy-policy
- type: Security
  url: https://docs.nvidia.com/sdgm/security-and-governance
- type: Compliance
  url: https://docs.nvidia.com/sdgm/security-and-governance
- type: TrustCenter
  url: security/kumoai-trust-center.yml
- type: VulnerabilityDisclosure
  url: security/kumoai-vulnerability-disclosure.yml
- type: DomainSecurity
  url: security/kumoai-domain-security.yml
- type: Authentication
  url: authentication/kumoai-authentication.yml
- type: Conventions
  url: conventions/kumoai-conventions.yml
- type: ErrorCatalog
  url: errors/kumoai-problem-types.yml
- type: RateLimits
  url: rate-limits/kumoai-rate-limits.yml
- type: Lifecycle
  url: lifecycle/kumoai-lifecycle.yml
- type: ChangeLog
  url: changelog/kumoai-changelog.yml
- type: Packages
  url: packages/kumoai-packages.yml
- type: SDKs
  url: packages/kumoai-packages.yml
- type: MCPServer
  url: mcp/kumoai-mcp.yml
- type: AgentSkill
  url: skills/_index.yml
- type: LLMsTxt
  url: llms/kumoai-llms.txt
- type: WellKnown
  url: well-known/kumoai-well-known.yml
- type: Conformance
  url: conformance/kumoai-conformance.yml
x-enrichment:
  date: '2026-07-19'
  status: backfilled
  pass: local-v1
  note: backfilled from .gitignore signal + verified work evidence