Pre-launch · Talking to 20 teams this month

Agent operations
for support.

Per-agent budgets, outcome attribution, and a prompt registry — on top of Zendesk or Intercom. No replacement required.

Book a 30-min call Join the design-partner list

If any of this sounds like your week,
we should talk.

We're not assuming these pains apply to you. We're talking to 20 teams this month to find out which ones actually do.

Your support agent is running on a human seat license, and you're not sure that's allowed.

A bad prompt burned hundreds of dollars in tokens overnight before anyone noticed.

You can't tell which agent run resolved which ticket, so you can't A/B test prompts against outcomes.

Your prompts live in three repos with no versioning.

Your agent silently degrades when Zendesk rate-limits it during traffic spikes.

When the agent escalates, the human inherits a thread with no context.

The agent operations layer
that sits next to your helpdesk.

Your helpdesk stays the system of record. Resolv runs the agents that touch it.

01

Identity & budgets

Per-agent token and cost budgets, rate limits, alerts. Hard enforcement when you ask for it.

02

Outcome attribution

Reconciles agent runs with the helpdesk's resolution state. Confidence tiers. Disputeable.

03

Prompt + eval registry

Versioned prompts and eval runs scored against actual ticket resolutions.

how it fits
your agent  ─►  Resolv  ─►  OpenAI / Anthropic
                  │
                  ├─►  Zendesk / Intercom  (read + write tickets)
                  │
                  └─►  resolutions ledger  ─►  attribution + billing

Why this, why now.

Zendesk and Intercom will ship their own MCP servers. That's the precondition for Resolv, not a threat to it.

The runtime layer wins whenever the substrate vendor's pricing model is misaligned with the workflow customers want. See Vercel on AWS, Datadog on CloudWatch, Metronome on Stripe — all built valuable companies on top of platforms that ship their own first-party versions of the same primitives.

A helpdesk vendor cannot ship per-agent budgets without cannibalizing per-seat pricing. So they won't. That's the gap.

Spring 2026 beta

Design partner program.

We're taking 3–5 design partners for our spring beta. Tight loop, direct line to the founder, real influence on what we build.

What you get

  • · Three months free during beta
  • · Direct Slack with the founder
  • · Real influence on the roadmap
  • · Named at launch

What we ask

  • · 30-min weekly call
  • · Real ticket volume
  • · Willingness to dispute attribution publicly
  • · Honest feedback, including the harsh kind

Form rendered by Tally in light mode for readability.

About.

I'm Bimal Singh. I'm building Resolv because I keep seeing the same gap on AI-forward support teams: production agents running on top of helpdesks that were never built for them.

If you run an AI agent in support — or you're the engineer responsible for it when it breaks — I'd like to talk. No deck, no slides, no pitch. Just a 30-minute call about how your team actually works.

Book a 30-min call