Your AI should get cheaper every time you use it.

Capture your AI work traces, automatically codify them into skills, serve them using in-house models, saving 80%+ on costs

Optimize your work
Operant console overview, drilling into a conversation savings ledger

The right model for each task. A fraction of the cost.

As Operant codifies your recurring work into skills and serves them with in-house models, the same tasks cost less every month — projected here, verified per call.

How the ledger works ->
month 1 · frontier baseline
100%
month 2 · first skills ratified
74%
month 3 · recurring work codified
46%
month 4 · served by in-house models
25%
steady state · −82%
18%

Illustrative projection of the same workload's bill as recurring tasks are codified into skills and served by in-house models. Savings are verified per call on the ledger, never assumed.

Point Operant at your AI agent.

Point Operant at your agent

One line in your agent's environment. No code change. Your agent keeps working exactly as before — through Operant.

gateway.operantlabs.com/#/routers/8
Operant
Overview
Conversations
Patterns
Routers
Skills
gateway · connected
Routers
operant/agent-runtimeLive
5 manual rules · default → claude-sonnet-4-5
$0.10
saved this month
operant/ops-assistantLive
1 learned rule · default → claude-sonnet-4-5
$0.00
saved this month
operant/standardLive
default → claude-sonnet-4-5
$0.00
saved this month
operant/standardLive
Copy endpoint snippet
ENDPOINT
https://gateway.operantlabs.com/c/<C>
Anthropic and OpenAI wire formats, both on this base URL.
DEFAULT ROUTE
claude-sonnet-4-5100%
.env — your agent
1# --- Point your agent here ---
2# was: api.anthropic.com
3
4# Stock Claude Code, grouped by conversation C:
5ANTHROPIC_BASE_URL=https://gateway.operantlabs.com/c/<C>
6# OpenAI SDKs (base URL includes /v1):
7OPENAI_BASE_URL=https://gateway.operantlabs.com/c/<C>/v1
one line per SDK · no code change
agent
$ claude "compile the SOC 2 evidence pack"
· operant/standard → claude-sonnet-4-5 · cache 94%
· skill v1 ready · 0 code changes

It learns the work

It watches how the agent works, learns from what it does, and trains smaller models to do the same job.

Operant
Overview
Conversations
Patterns
Routers
Skills
gateway · connected
Patterns
demo-compliance-5
The user wants to compile the June SOC 2 monthly evidence pack and evidence two emergency changes that shipped without pre-approval.
4
$0.02
demo-compliance-5
The user wants to generate the quarterly access review report and chase down unreviewed accounts.
5
$0.02
demo-compliance-5
The user wants to assemble the May SOC 2 evidence pack and close the access-review gaps carried over from April.
4
$0.02
demo-compliance-5
The user wants to assemble the July SOC 2 evidence pack and reconcile the incident log against the on-call handover notes.
4
$0.02
demo-compliance-5
The user wants to compile the November SOC 2 monthly evidence pack and reconcile a gap in the deploy log.
6
$0.01
demo-compliance-5
The user wants to assemble the August SOC 2 evidence pack and document the access recertification exceptions for terminated contractors.
4
$0.01
demo-compliance-5
The user wants to produce a customer-facing access-controls report for an enterprise security review.
5
$0.0093
demo-compliance-5
The user wants to assemble the December SOC 2 evidence pack covering the change-freeze period and the year-end access recertification.
5
$0.0087
localhost:5173/#/conversations/demo-compliance-5

A new model must earn the job

Operant keeps testing them. A new model only takes over when it is better, cheaper, faster, or safer.

Operant
Overview
Conversations
Patterns
Routers
Skills
gateway · connected
Patterns
Manage Customer Password Access
Users need to reset, unlock, or troubleshoot password access for customers who are locked out, locked in, or facing authentication barriers due to forgotten passwords, security incidents, technical issues, or account state changes.
12 conversations$0.03 · 4.9% of spendlast seen 9h ago$0.01 compressible / mo · 19% cache hits · 1 stale tool tokens/call
Skill
Scribe drafts the skill from this pattern's captured traces; you review and approve it into the route/eval flow.
▾ What happens on Learn skill?
Scribe selects diverse, successful member conversations (reserving a disjoint holdout for the eval gate), bundles their traces, and drives a harness session seeded with the learning skills. The drafted SKILL.md lands in the registry as a draft version with full provenance — nothing reaches traffic without your approval, the eval gate, and route ratification. Qualifying patterns re-learn automatically on the heartbeat when they drift.
Learning — 24s
Member conversations
CONVERSATION
CALLS
SPEND
demo-pwreset-4
The user wants to reset a customer's password and MFA enrollment after the customer lost their phone.
2
$0.0060
demo-pwreset-9
The user wants to force a password reset across a team whose shared credential was exposed.
2
$0.0047
demo-pwreset-8
The user wants to unlock an account that keeps re-locking after repeated failed sign-ins.
2
$0.0044
Skill registry
Versioned instruction units distilled from a pattern's exemplars — injected as a system-prompt append when their route is active.
Compile SOC 2 Evidence Packs
Gather and organize monthly or quarterly compliance evidence for SOC 2 audits: access reviews, incident logs, change management records, and exception documentation with remediation tracking.
v1ready
New version
Manage Customer Password Access
Reset, unlock, or troubleshoot password and MFA access for customers facing authentication barriers from forgotten credentials, lost devices, lockouts, or security incidents.
v1draft
New version
Manage Customer Password Access
manage-customer-password-access · v1draftManage Customer Password Access
Reject
Approve to ready
New version
Reset, unlock, or troubleshoot password and MFA access for customers facing authentication barriers from forgotten credentials, lost devices, lockouts, or security incidents.
LEARNED BY SCRIBE
claude-sonnet-4-5$0.09 learning cost9 agentic turnsbundle c920c100ffrun 2-1787294645367
▸ Taught by 1 conversation (a disjoint holdout stays reserved for the eval gate)
Assess the access issue
Determine what the customer needs: password reset only, MFA reset only, or both. Check for existing tickets or recent incidents using search_tickets with the customer identifier or relevant keywords.
Password-only resets are standard-risk. MFA resets or combined password+MFA resets are higher-risk because they bypass multi-factor protections. Account lockouts may be automatic (failed login attempts) or manual (security-triggered); identify which.
Verification requirements
Match verification rigor to the risk level:
Standard password reset: Basic identity confirmation (email verification, security questions, or account details like billing address).
MFA reset or combined reset: Higher bar required. Acceptable methods:
Government-issued ID verification over a recorded video call, OR
Two out-of-band factors: billing address + last four digits of payment method + a recent transaction detail.
localhost:5173/#/patterns/2

You own the models.

FAQ

A few things people ask first.

No. Operant speaks the Anthropic and OpenAI wires. You change one base URL. You keep your prompts, tools, and SDKs. Address a router as the model, and the gateway chooses for each turn.

Yours. Store a provider key, encrypted at rest. Mint an Operant key for each agent. The gateway strips the Operant key, injects yours at forward time, and stamps every call with the key that sent it.

You decide. Observe-only is the default and changes nothing on the wire. Capture is tiered. The trace shows both what the app said and what Operant sent to the provider. You can revoke in a minute. Point the base URL back.

Nothing changes without proof. Operant replays held-out work through the candidate and scores it against the frontier model’s own answers. It takes live traffic only when it matches or beats that baseline and you ratify. One click demotes it, and demote is never gated.

Every dollar of agent spend, priced —
and the waste removed.