In early access — working directly with design partners

Universal Orchestrator for Enterprise Agents.

Culvii kit abstracts the identity, controls, and auditability for developers put agents in production in record time — and the operations team to keep them running. Two stakeholders, one mission: make Enterprise Agents safe.

Working with engineering teams at JLL, Visot, and other enterprises in regulated industries.
verified audit · live merkle · f4a2…
agentdid:culvii:jll-prod:lease-extractor-a3f9b2
tenantjll-prod · multi-tenant-isolated
workflowlease-review-q2 · step 4 of 6
capabilityread:leases · scope=q2-portfolio
modelclaude-3-5-sonnet · anthropic
actiontool.call(extractClauses) · 840 ms
signatureed25519 · 5e3c…c071 · verified ✓
hitlapproved · legal@jll · +0.4 s
timestamp2026-04-21T14:32:07Z
event#142 of 142 · hash-chained
Trusted by engineering teams at
JLL Visot + more soon

Build them. Run them.

Enterprise AI agents need two things most companies underestimate: a governance-first SDK to build them right, and an operations team to keep them healthy. We offer both.

SDK @culvii/kit · v0.0.1

Build agents your security team will approve.

The TypeScript and Python SDK for enterprise AI agents. Cryptographic identity, capability governance, HITL controls, and immutable audit trails — out of the box.

  • Cryptographic agent identity (DID + Ed25519)
  • Capability registry with explicit grants
  • Hash-chain + Merkle Tree audit trails
  • Human-in-the-loop checkpoints
  • TypeScript + Python SDKs
Managed Service micro · mega · ultra

A dedicated AgentOps team, without the headcount.

Embedded specialists who operate, monitor, and continuously improve your AI agent infrastructure. Enterprise Operations consultants, AgentOps Engineers, and Runtime Experience Engineers — fractional or dedicated.

  • Enterprise Operations consultant (strategy & governance)
  • AgentOps Engineer (deployment & incident response)
  • Runtime Experience Engineer (reliability & observability)
  • 24-hour turnaround SLA from Micro tier up
  • Micro, Mega, and Ultra tiers to match your scale

The LLMs are ready.
The enterprises aren't.

Every enterprise engineering team is being asked to ship AI agents. Almost none of them are. The blocker isn't capability — it's that the typical agent framework gives security teams nothing to hold onto. No identity. No access controls. No audit trail that can withstand a compliance review.

We've watched this play out at enterprises across financial services, real estate, and regulated SaaS: engineering teams build something genuinely useful, then stall at the security review. Culvii ends that standoff — with infrastructure on one side, and operations on the other.

Fig. 01 · the enterprise agent stall source · culvii field research, 2026
0
leading agent frameworks treat identity as a first-class concern. LangChain, CrewAI, AutoGen — we checked.
6+ mo
typical time for an enterprise platform team to build agent governance infra from scratch.
2 weeks
typical time to integrate Culvii Kit into an existing enterprise agent stack. See case studies →

Quotes from the people shipping on Culvii.

"
Quote pending approval.
J
Senior Engineer · JLL
"
Quote pending approval.
V
Platform Lead · Visot
"
Quote pending approval.
F
Head of Platform · Enterprise FinServ
Culvii Kit

You wouldn't deploy a microservice without authentication.

Culvii Kit assigns every AI agent a Decentralized Identifier (DID) backed by Ed25519 keypairs. Every action an agent takes is signed with its private key.

Your security team can verify the identity of any agent, anywhere in the organization — and revoke it instantly if something goes wrong. Compromise one agent; don't compromise them all.

agent identity · issued ed25519 · revocable
diddid:culvii:acme-prod:invoice-triager-0c98f2a
issueracme-prod.registry.culvii.com
public keyed25519:5e3c f8b1 d240 9a7e c071 …
created2026-04-18T09:14:02Z
status● active · rotation scheduled 2026-07-18
scopetenant=acme-prod · cross-tenant=false
actions signed · 12,847 tamper-evident ✓
AgentOps

Most enterprise AI projects don't fail at launch. They fail at operations.

A governance SDK is necessary but not sufficient. Someone still has to operate your agent fleet: deploy updates, respond to incidents, monitor runtime behavior, align with stakeholders on governance decisions, and continuously improve performance.

AgentOps gives you that team — fractional or dedicated, embedded in your workflow, named people with dedicated context on your environment. Not a rotating support queue.

embedded team · acme-prod weekly ops review · thu 14:00
EO
Enterprise Operations Consultant
strategy · governance · exec reporting
fractional
AO
AgentOps Engineer
deployment · incidents · ci/cd
dedicated
RX
Runtime Experience Engineer
slo · reliability · observability
dedicated
sla · 24h turnaround tier · mega
Enterprise

Built for regulated industries.

Culvii was designed from day one for organizations operating in compliance-sensitive environments — financial services, real estate, legal, and healthcare.

Multi-tenant isolation by design
Per-tenant key scoping, capability namespacing, data isolation from the ground up.
SSO + SCIM provisioning
Okta, Azure AD, SAML 2.0. Provision and deprovision without opening a ticket.
On-premise & BYOC deployment
Run Culvii in your own VPC or air-gapped environment. Your network, your keys.
Dedicated support
Named TAM, 24/7 critical support, guaranteed response SLAs — not a queue.
Compliance: SOC 2 Type II · in progress GDPR HIPAA ISO 27001 · roadmap
Talk to our enterprise team

Public changelog. Every sprint.

View all
Apr 21, 2026 feature Actor primitive in active development. Core identity-bearing execution unit with Ed25519 keypair generation.
Apr 20, 2026 dx Docker Compose streamlined. Single docker-compose up. No config.
Apr 14, 2026 feature Workflow primitive in development. Sequential, parallel, and conditional paths with HITL checkpoint support.
Apr 02, 2026 milestone v0.0.1 shipped. First public release of the @culvii/kit monorepo.

Give your CISO something they can sign off on.

Whether you want to build governed agents yourself with Culvii Kit or have us run them with AgentOps — we'd rather start with a conversation than a pitch.

No SDR. No deck. You'll speak with a founder.