Pitch · 2026
Six AI agents covering Treasury, Capital, IR, Books, AP, and AR. On-chain-first. Agent-driven. Non-custodial. The wallet is the books — and the agents do the work.
The bet
Token component, stablecoin treasury, global team paid in USDC, on-chain investors. Capital.xyz serves the prior generation — Delaware C-corps with Mercury accounts and dollar-denominated cap tables. They're not going to retrofit chain-native primitives because their customer doesn't want it. We're building for the cohort they're not serving.
The unlock
The agent layer removes the "but we're too small for a CFO" problem. A two-person startup with a token can run their entire money / equity / legal operation through agents that cost ~$500/mo — instead of hiring a fractional CFO + a lawyer + a fund admin.
The cohort
The wedge
The category
e.g. capital.xyz · Carta · Pulley
What we ship
On-chain-first · agent-driven · non-custodial
Why now
Foundations + protocols + DePIN networks now manage 7-9 figures of treasury without a finance team. The pain is real and the budget exists.
Safe Modules + Bridge.xyz + Resend Inbound + Coingecko + Across mean every primitive an AI back office needs is composable, audited, and cheap.
Claude Sonnet 4.6 + Llama 4 + GPT-OSS run reliable structured-output workflows for under a dollar. The agent-coworker pattern stopped being a demo and became infrastructure.
The team
Foundation agents do the bookkeeping that everyone has — but ours feeds the headline agents. Headline agents do things only crypto-native autonomy enables.
Multi-chain reconciliation, Safe-Module-driven yield deployment, stablecoin rebalancing, runway alerts. You set the policy; the module executes.
Cap table for equity AND tokens, SAFT/SAFE drafting, on-chain vesting contract deployment that mirrors legal terms. Carta can't do on-chain vesting; we do.
Monthly investor updates from real numbers, board pack assembly, runway/burn reports, fundraising pipeline + data-room curation, term-sheet status tracking.
On-chain ↔ off-chain reconciliation, Xero/QBO sync, monthly close packet. Bridges DeFi positions, cross-chain transfers, vesting events natively.
Inbox-as-source-of-truth for bills. Counterparty awareness, sanctions screening, payment proposals signed from your wallet.
Invoice generation, stablecoin checkout via Bridge.xyz, payment reconciliation back to Books.
Compliance Officer + Tax Accountant ship in v1.5. Treasury Risk, Fundraising Ops, Procurement, DAO Ops join the catalog as design partners ask. Every agent is a named hire — admins build the team via a hire flow, not a feature toggle.
Build vs buy
Claude, Paperclip, OpenAI Assistants — they all wire models to tools and let you hand-roll the workflow. That's the easy 5%. The other 95% is the on-chain plumbing, the non-custodial signing pipeline, and a back-office team that talks to itself.
Embedded wallet + Safe Module + ExecutionSuccess / Failure event parsing wired into an action ledger. Generic orchestrators call tools; they don't run a propose → human approves → wallet signs → on-chain receipt loop without taking custody. Without this you're either a custodial RIA or you ship buttons that go nowhere.
Multi-chain RPC manager (Base, Ethereum, Arbitrum, Optimism, Polygon), USDC reconciliation, vesting event reads, DeFi position parsers, Coingecko pricing, Bridge.xyz fiat rails, sanctions screening. Generic agents read your inbox; ours read your wallet. The integration tax is six engineering months — and stays a maintenance burden forever.
One heartbeat orchestrator, run-id mutation discipline, wakeup queue with coalescing, persistent per-agent state, cross-agent commentary (AP confirms a Helios bill needs EUR; Treasury proposes a USDC → EURC bridge in the same heartbeat). Off-the-shelf orchestrators run one agent per query. We run a back office.
The moat
The wallet IS the books. Multi-chain treasury, token-grant vesting, on-chain SAFTs are native primitives, not translated objects. Fiat-first competitors fundamentally can't do this well even with AI bolted on — they'd need to rebuild their data layer.
Every back-office function is a named agent that drafts → human approves → on-chain execution. Different mental model from dashboards-and-buttons. The team metaphor is real — each agent has a job description, scope, and the operator can hire/fire them.
Treasury rules, comp policy, counterparty policy aren't dashboards — they're smart contracts that enforce themselves. Sign once at policy time; contract executes within bounds; every action auditable. Switching means redeploying every contract. The moat compounds with usage.
What we are not
Bridge.xyz handles fiat banking + virtual cards via their existing partner program. We integrate, not custody.
Stablecoin payments via Thirdweb / Coinbase Commerce. Standard integrations.
Capital agent prepares SAFTs / SAFEs and deploys vesting contracts. Securities transactions happen between customer and investor directly.
Every action is signed by your wallet (or your pre-authorized Safe Module within rules you set). We hold zero customer assets at all times.
We do not exercise discretion. You set the policy; agents draft within it; you (or your Safe Module per your rules) is the deciding actor.
Pricing
Per-action and per-seat. Above your tier's included quota, additional agent actions meter at a flat rate. The pricing model is the same shape as Vercel, Supabase, Linear — usage-aware SaaS, not asset-under-management investment-manager pricing.
1 wallet · AP + Books · read-only
1 user · 3 wallets · 3 hires · 50 bills/mo
5 users · ∞ wallets · 6 hires · 1k actions/mo
15 users · 3 entities · v1.5 catalog · 5k actions/mo
Unlimited · dedicated onboarding · SLA
For investors
Every Safe Module a customer deploys, every counterparty alias the agents learn, every vesting contract Capital ships — those live in the customer's on-chain state. Switching means redeploying contracts. Every quarter we're harder to leave.
QuickBooks / Carta / Brex can ship AI tomorrow. They cannot ship on-chain vesting deployment, Safe-Module-enforced policy, or wallet-as-the-books reconciliation. The technical moat is real and recent (Safe Modules went mainstream in 2024).
Crypto-native startups at Series A+ scale (foundations, DAOs, protocols, DePIN networks) cluster in known geos and accelerators. Distribution is tractable; willingness to pay is high — we replace a $150K/yr controller hire.
v1 ships six agents. v2 grows the catalog (Compliance, Tax, Treasury Risk, IR-extension agents, etc). Each new agent is a new pricing lever and a new reason to upgrade. The unit economics improve over the customer's lifecycle.
The lighthouse
v2 exposes our agents as MCP-callable services. Customer-side AI agents (the operator's own AI assistants) call our Treasury and Capital agents directly. The propose / approve / sign spine works whether the proposer is a Softmax agent or an external AI signing under policy. We're already building toward it; it's in the schema.
What now
We're building the long-form story. Reach out for the deck + traction snapshot.