Full site context for agents: https://tansohq.com/llms.txt

Tanso vs. Orb, Stigg, Schematic, and Metronome

Four strong hosted billing platforms, one AI usage runtime (Limitr, by Formata), and an open-source engine with a different job. They give you a console to run your billing. Tanso runs the billing and also gives your customers' agents a way in: readable pricing, signup, payments, spend caps.

Capability Tanso Orb Stigg Schematic Metronome Limitr
Open source / self-host open source, self-hosted: one application plus Postgres Proprietary, hosted only Proprietary; hosted or BYOC in your own VPC Proprietary, hosted, built on Stripe Proprietary, hosted only Apache-2.0 engine on GitHub; runs in your process (WASM), plus a cloud
What the vendor charges you Nothing. No vendor bill, no per-event fees Custom pricing on every tier; billed on invoice value plus ingested events Free tier, then $399/mo plus per-entity and per-event fees Free under $5K/mo billing volume; Growth $400/mo; Enterprise custom 0.8% of billed amount plus $0.04 per 1K events; custom above that No public pricing
Credit wallet Yes. Pools, grants, expiry, rollover, purchased top-ups Yes. Prepaid credits with real-time balance Yes. Stacked grants and consumption formulas Yes. Credit burndown from prepaid pools Yes. Credits plus prepaid and postpaid commits on contracts Credits defined in policy config; wallet details not documented
Reprice actions without a deploy Yes. Versioned, effective-dated credits-per-action table, edited in the console No credits-per-action weight table documented Consumption formulas in the console; no dated weight versions documented Plan limits editable; no credits-per-action weight table documented Contract amendments; no credits-per-action weight table documented Policy config, not code; dated versions not documented
Price-per-credit book Yes. Effective-dated book; grants stamp the book price, negotiated overrides win Configured per plan; no dated price book documented Prices stay in your billing provider Prices stay in Stripe List pricing with per-contract overrides Not documented
Cost and margin per customer Yes. Cost rules per feature; margin per customer in the same ledger Claimed on marketing (/solutions/ai); no cost ledger documented Not documented Not documented Not documented Yes. Per-call margin with vendor, customer, and feature attribution
Internal AI spend (Anthropic, OpenAI, Cursor, Copilot, LiteLLM admin APIs) and Feature P&L Yes Not documented Not documented Not documented Not documented Not documented
Customer-facing forecast UI Yes. Burndown API plus a React widget; depletion date per pool No forecast UI documented Paywall and portal widgets; no burn forecast documented Embeddable plan and usage components; no burn forecast documented Real-time balance APIs; no forecast UI documented Not documented
Machine-readable public pricing Yes. pricing.json per catalog: one GET, no auth No No No. Their own docs are agent-readable (llms.txt), your pricing is not No No
Agent signup and payments (402 flow) Yes. One-call signup issues a scoped key; 402 with a pollable checkout session id No No No No. Stripe's agentic commerce work sits at the payments layer, not in Metronome No
MCP surface Operator tools plus curated customer tools; spend needs explicit confirmation Operator-facing agent connector for your Orb account Yes, operator-facing: manages your Stigg environment Yes, operator-facing: manages your Schematic account None documented; Stripe's MCP covers payments None documented
Spend caps for agent buyers Yes. Per-account top-up cap enforced before money moves No No No No Caps enforced at call time; no agent-buyer flow documented

Facts checked September 2, 2026 against each vendor's public docs and pricing pages. "Not documented" means we could not find it published, not that it does not exist. Corrections: kat@tansohq.com

The rows that are not a feature race

Every platform here gives you a console, an API, and in most cases an MCP server for managing your own account. That is operator tooling. The last four rows are a different axis: whether your customers' agents get a surface of their own. An agent shopping on a customer's behalf needs pricing it can read without a login, a signup call that returns a scoped key, a payment path that survives a 402 and can be resumed by id, and a cap on what it can spend. None of the other platforms documents any of that today. Tanso ships all of it, and the wallet, weight table, and price book are what make it safe to turn on.

Different tools for different jobs

Orb: usage-based billing at serious event volume with a finance workflow around it. Strengths: deep metering (250K+ events/sec, custom SQL metrics), prepaid credits, Adyen behind it since mid-2026. Weaknesses: custom pricing only, hosted only, no surface for your customers' agents.

Stigg: entitlements and credits as a layer over the billing provider you already run. Strengths: fast entitlement checks (documented sub-100ms P95), no-code console, BYOC and air-gapped options. Weaknesses: not open source, per-entity and per-event fees, MCP manages your Stigg account rather than serving your buyers' agents.

Schematic: Stripe teams that want plans, entitlements, and credit burndown without hard-coding checks. Strengths: Stripe-native sync, embeddable components, free under $5K/mo billing volume then a flat $400/mo, agent-readable docs. Weaknesses: requires Stripe, hosted only, no cost or margin ledger documented.

Metronome: enterprise consumption contracts: commits, amendments, and marketplaces at proven scale (OpenAI, Anthropic, NVIDIA). A Stripe company since January 2026. Weaknesses: 0.8% of billed amount plus event fees, hosted only, agentic commerce lives at the Stripe payments layer.

Paid.ai: companies selling AI agents to enterprises, priced on outcomes and proven with ROI receipts. Strengths: outcome and per-agent pricing, margin tracking, SOC 2 / GDPR / HIPAA / ISO 27001. Weaknesses: hosted only, not open source, and it bills the humans who buy your agents; it gives no surface to agents buying from you.

Limitr: AI products that want usage gated and margin measured at call time, on top of whatever bills the customer. Strengths: per-call margin with vendor, customer, and feature attribution, limits enforced in your process before the vendor call, Apache-2.0 engine on GitHub with a TypeScript SDK (@formata/limitr). Weaknesses: no public pricing, not a billing system (invoicing stays elsewhere), no agent-buyer flow or MCP documented.

Tanso: products that want their features usable by their customers' agents, with the billing machinery that makes that safe to turn on. Strengths: public pricing.json, one-call agent signup, 402 with a pollable checkout session id, spend caps, curated MCP tools, versioned weight table and price book changeable without a deploy, margin per customer in one ledger, open source self-hosted with no vendor bill. Weaknesses: newest here, smallest feature surface, self-host only, no SLA, no SOC 2, Stripe-only payments.

Links

  • Tanso source · docs
  • Limitr · source
  • Tanso vs. Lago vs. Autumn (open-source peers)
  • HTML version of this page