Today every SaaS purchase takes weeks and a procurement team. We turn it into one API call. Vendor evaluated, policy enforced, an AI agent fires the negotiation — your code keeps moving.
Every SaaS purchase still flows through a chain of humans, Slack threads, spreadsheets, and approval emails. The cost is hidden in cycle time, missed savings, and engineers context-switching to fight finance. We're shipping the API alongside our first 10 design partners — be txn_0001.
Procurement-as-an-API isn't a feature. It's what happens when AI agents become reliable enough to negotiate on your behalf, dev-led adoption reaches finance, and SaaS sprawl breaks every existing tool.
Multi-turn vendor conversations with structured concession ladders are tractable on frontier LLMs (GPT, Claude). For the first time, "the agent ran the negotiation" survives audit and closes against your floor — not the vendor's anchor.
Ramp, Mercury, Stripe Tax — the next-gen finance stack is API-first. Procurement is the last function that still ships as a workflow tool. Devs are ready for the primitive.
The median company runs 250+ SaaS contracts. No procurement team can renegotiate them all. The long tail leaks 10–20% — automation is the only path that scales.
Same flow on the landing page, in the demo, and in production. Receive, evaluate, decide, fire the webhook — your code keeps moving while the negotiation runs async.
One endpoint. Send vendor, amount, category, term. Anything from a $200 SaaS seat to a $2M services contract.
Policy, budget, third-party risk, vendor history — checked simultaneously against your config. Each returns a signed pass/fail rationale.
The engine returns a signed decision with full rationale. Your code keeps moving. No human in the synchronous path.
Agent opens the vendor thread, runs the concession ladder, captures the signed counter, and posts to your webhook with the final contract.
Four capabilities ship the moment you authenticate. No onboarding. No rules engine to configure first. Sensible defaults that adapt to your spend as the engine learns.
Policy + budget evaluation in milliseconds. Per-category caps, per-owner thresholds, fiscal-period awareness. Every decision returns a signed rationale.
Vendor comms handled automatically. Email-based, structured concession ladder, walk-away-aware. Closes against your floor, not the vendor's anchor.
SOC 2, DPA, TPRM, DPIA — checked against vendor record. Flags missing artifacts and routes to the right reviewer when needed.
Every decision recorded, every negotiation indexed. The engine learns your tolerance and improves over time.
New categories live or die by what they refuse to be. Here's where we draw the line.
If your job-to-be-done is "route a request through 7 approvers," buy Zip or Coupa. Nexum is the API that runs after the workflow decides to buy.
No humans in our path. If you want a negotiator on retainer for $500K+ deals, hire Vendr. We're the primitive for the long tail your broker won't touch.
Dashboards observe. We act. Outcome tracking is a side-effect — the product is the decision the engine made and the negotiation an AI agent ran while you were doing something else.
Humans negotiate your largest contracts deal-by-deal. Right call when one broker pays for themselves on a $500K+ deal. Not API-first, not embeddable.
One endpoint. Decision in seconds. Negotiation async. Plug it into your finance stack, ERP, or your product. Built for the long tail of every other vendor decision.
UI for routing requests through approvers. Right call when policy is human judgment by design. Not API-first, not embeddable.
Not paying customers. We're talking to design partners now and shipping the API alongside their first transaction. If you want to be transaction txn_0001, this is the moment — pricing, scope, and roadmap are still being shaped by the first cohort.
The negotiation agent uses frontier models (GPT, Claude, swap-able) for the multi-turn vendor conversation, yes. But the decision engine is deterministic — policy, budget, and TPRM checks are code, not prompts. You get a signed rationale you can audit, not a vibes-based outcome.
Config-driven. You define category caps, owner thresholds, fiscal-period rules, and vendor allowlists in a JSON config or the dashboard. Override a decision and the engine learns the override pattern. Generic defaults ship for fast onboarding.
Every decision returns a signed rationale. Full request/response/check log per transaction at /v1/transactions/:id/audit. Replayable. We're targeting SOC 2 Type II for GA.
Webhook-first — anything HTTPS works on day one. Native ERP/finance integrations land with design partners — we build the adapters you need, not the ones we guess at.
Returns "decision": "escalated" with reason and a routing target (owner email or Slack channel). You own the human-loop branch — we hand off cleanly. Try the $2M scenario in the demo.
Not yet. We're holding pricing until the first cohort tells us what the value is. Design partners get pricing locked at launch.
Design partners onboarding now. We ship the API alongside your first real transaction — pricing, scope, and integration shaped by the first cohort. Reply within 24 hours.
Or email alex.chang@nexumlabs.ai directly. We respond within 24 hours.