nable capabilities
July 20, 2026 ยท View on GitHub
Everything in this repository is open source under Apache-2.0. nable is a local-first cost engine that runs on your machine and exposes the same intelligence three ways: a terminal CLI, an MCP server for Claude/Cursor, and a local web dashboard. Your bill never leaves your machine.
At a glance: 194 MCP tools, ~40 connectors, ~48 CLI commands. Read-only by default. Credentials stay in your OS keyring, never routed through any model.
What is nable?
nable is an open-source, local-first cloud and AI cost tool. It finds recoverable spend across AWS, Azure, GCP, Kubernetes, and 15+ SaaS and AI providers, shows what your LLM and GPU usage actually costs, and proposes each fix as a pull request you approve. It runs on your machine as a terminal CLI and as an MCP server for Claude and Cursor. It is free to start, read-only by default, and your bill never leaves your machine.
It is a fit for a DevOps or platform engineer who gets handed a rising cloud and AI bill and needs to cut it, from the terminal, without shipping billing data to a third party.
The one command
uvx nable scan
Finds recoverable spend on your AWS account in under a minute, from the free
AWS APIs, so it never puts a charge on your bill. --spend adds a Cost Explorer
breakdown (~$0.02, disclosed first), --json for CI, --demo for sample data
with no account at all.
MCP tools (194)
The same engine that answers in your terminal answers in Claude, Cursor, Windsurf, and VS Code. Ask in plain terms; nable picks the tool.
| Area | Tools | What you get |
|---|---|---|
| AWS waste + scans | 34 | idle NAT gateways, unattached EBS, oversized RDS/EC2, old snapshots, EIPs, Graviton and spot migration, CloudWatch/S3/Lambda/ECS/ECR waste |
| Cost queries | 27 | spend by service, team, account, region; trends; forecasts; cross-source totals; FOCUS; top drivers; "why did the bill move" |
| Meta / RBAC / connectors | 19 | connected providers, connector health, API keys, capabilities, identity |
| Attribution | 13 | tag rules, team and label allocation, showback, commitment coverage by tag |
| LLM / AI cost | 12 | spend by model, unit economics, token forecasting, AI spend monitor, billing blind spots, model routing |
| Notifications | 13 | Slack, Teams, Notion, n8n, weekly digests and insights |
| Kubernetes | 11 | namespace, workload, and cluster cost; efficiency scorecards; waste tickets |
| Rightsizing + recommendations | 10 | EC2/RDS/ECS rightsizing, idle cleanup, remediation PRs, a learning loop |
| Commitments | 8 | Savings Plan and Reservation analysis, coverage, RI waste |
| Tickets | 8 | auto-create issues in Jira, Linear, GitHub from anomalies and waste |
| Azure | 8 | Cost Management, Advisor, VM rightsizing, reservations, budgets |
| AWS core | 7 | connect, accounts, org rollup, CUR pipeline |
| Budgets | 5 | set, sync from YAML, two-tier enforcement (warn / block) |
| Anomalies / GCP / Databricks / Forecast | 3-3-3-1 | anomaly detection and alerting, GCP waste, DBU consumption, spend forecasting |
Connectors (~40)
Clouds: AWS (plus Organizations, CUR, detailed billing), Azure, GCP, Vertex. Kubernetes: native, Kubecost, OpenCost, Helm releases, GPU infrastructure. AI / LLM: OpenAI, Anthropic, Amazon Bedrock, OpenRouter, LiteLLM, Modal, Together, Replicate, Cohere, Mistral, Langfuse. Data platforms: Databricks, Snowflake, MongoDB Atlas. SaaS: Datadog, New Relic, Cloudflare, Twilio, Vercel, Stripe. Infrastructure as code: Terraform plan cost, Terraform estimate.
Connect AWS or GCP by SSO login or a CLI profile and nable stores no secret at all, it just references your existing login. Only keys you paste directly are encrypted in your OS keyring.
CLI (~48 commands)
get answers scan
start here welcome connect doctor tools serve upgrade
clouds aws aws-cur azure gcp
ai providers openai anthropic openrouter litellm modal together
replicate cohere mistral
saas datadog newrelic databricks snowflake mongodb twilio
cloudflare vercel langfuse
alerts slack teams notion n8n
editor+agents claude guard agents
account login logout license license-status credits
advanced config vault profile sso iam-template infra
finops connect scans the machine for provider credentials and connects
everything in one keystroke. finops serve starts a local web dashboard the
whole team can view in a browser.
The agent team
Propose-only, never auto-execute. The code is open; these features run on Pro.
- Budget Guard checks infra actions against your cost policy before they run.
- Savings Analyst proposes each fix as a pull request you approve.
- The Ledger tracks realized savings, measured against your actual next bill.
Plus finops guard (a cost guardrail your AI agents check before spending) and a
estimate_change_cost preflight on Terraform and Helm diffs.
Free vs Pro
Breadth is free on purpose: one tool across every cloud, SaaS, and AI provider.
Free forever: every cost query, anomaly detection, all waste and rightsizing
findings, nable scan, every connector, Slack and Teams alerts, PR comments.
Pro: the agent team, ticket auto-creation, scheduled email digests, commitment recommendations, org reports.
Running as an MCP server inside Claude or Cursor uses your existing membership as
the model. No Anthropic API key, no per-token cost. nable scan is deterministic
and uses no model at all.
How nable compares
| nable | AWS Cost Explorer / Cost Optimization Hub | Infracost | Vantage / CloudHealth | |
|---|---|---|---|---|
| Open source | Yes (Apache-2.0) | No | Partly | No |
| Runs locally, no data egress | Yes | N/A (in-console) | Yes | No (SaaS) |
| Cross-cloud (AWS + Azure + GCP) | Yes | AWS only | Pre-deploy IaC | Yes |
| AI / LLM / GPU cost | Yes | No | No | Limited |
| Terminal + CI | Yes | Console only | CI (PR comments) | Dashboard |
| Works inside Claude / Cursor (MCP) | Yes | No | No | No |
| Proposes fixes as PRs | Yes | No | No | No |
| Price | Free to start | Free (AWS-native) | Free tier | Paid |
nable is the open-source, cross-cloud option that also covers AI and GPU spend and runs on your own machine. Single-cloud waste (idle EBS, oversized RDS) is available free from the cloud's own tools; nable's difference is unifying every cloud, SaaS, and AI provider in one place, in the terminal, with fixes you approve.
FAQ
Is nable free? Yes. The terminal scan, every cost query, anomaly detection, all waste and rightsizing findings, and every connector are free forever. Ticket auto-creation, scheduled digests, commitment recommendations, and the agent team are Pro.
Does my billing data leave my machine? No. nable is local-first and read-only by default. It reads your cost and usage data on your machine and never uploads it. The no-egress claim is something you can read in the source, not take on faith.
What clouds and providers does nable support? AWS, Azure, GCP, and Vertex; Kubernetes (Kubecost, OpenCost, Helm, GPU infra); AI and LLM providers (OpenAI, Anthropic, Amazon Bedrock, OpenRouter, LiteLLM, Modal, Together, Replicate, Cohere, Mistral, Langfuse); data platforms (Databricks, Snowflake, MongoDB Atlas); and SaaS (Datadog, New Relic, Cloudflare, Twilio, Vercel, Stripe).
How is nable different from AWS Cost Explorer?
Cost Explorer is AWS-only and lives in the console. nable is cross-cloud, runs in
your terminal and in Claude/Cursor, covers AI and GPU spend that no cloud console
shows, and proposes fixes as pull requests. The nable scan default also makes
zero paid API calls, so it never charges your AWS bill to show you your own costs.
Is there an open-source alternative to Vantage or CloudHealth? nable is an open-source (Apache-2.0), local-first alternative for cost queries, waste detection, rightsizing, and AI/GPU cost. It runs on your machine instead of a hosted SaaS, so your billing data stays local.
How do I track LLM and GPU cost? Connect your AI providers and ask nable for spend by model, unit economics, token forecasting, and AI billing blind spots, across OpenAI, Anthropic, Bedrock, Modal, and more, in one view.
Do I need an API key or a credit card?
No. uvx nable scan runs with the AWS credentials already on your machine. As an
MCP server, it uses your existing Claude or Cursor membership as the model, with
no Anthropic API key and no per-token cost.
What is nable's MCP server? An implementation of the Model Context Protocol that exposes 194 cost tools to Claude, Cursor, Windsurf, and VS Code, so you can ask about your cloud and AI spend in chat and get answers priced on your real rates.
Trust
- Local-first. Everything runs on your machine. Your bill never leaves it.
- No-egress. A claim you can read in the source, not take on faith.
- Read-only by default. nable never changes your cloud on its own.
- No credentials through the model. Secrets stay in the OS keyring or a 0600 file; no MCP tool takes a secret as an argument.
- Apache-2.0. Open and auditable. A hosted platform is available for teams.
See getnable.com for docs and the hosted platform.