Limetry Cloud — organization policy checks, approvals, and audit for teams on Limetry
Organization-wide governance,
not one shared master key.
Centralize policy authoring, scoped agent identities, and team permissions so engineering teams ship autonomous agents safely without sharing root secrets.
25k
evaluates / month on Free
14 / 90 / 365
audit days on Free · Pro · Business
250k
evaluates / month on Pro
Unlimited
decisions on Business
evaluate.ts
import { createRemoteEngine } from "@limetry/sdk"
const engine = createRemoteEngine({
baseUrl: process.env.LIMETRY_BASE_URL,
apiKey: process.env.LIMETRY_API_KEY,
})
const decision = await engine.evaluateAction({
intent_id: crypto.randomUUID(),
policy_id: policyId,
agent_id: "ci-bot",
action_type: "deploy",
resource: "github.com/acme/api@abc123",
issued_at: new Date().toISOString(),
})
if (!decision.ok || !decision.approved) {
throw new Error((decision.reasons ?? ["denied"]).join("; "))
}
Organization controls built in
Shared policies, scoped agent tokens, privacy-safe audit, and an approval inbox on the same action intent the open-source Limetry SDK already uses.
Operators in the loop
When policy needs a human, approvals land in the portal or mobile inbox — with signed receipts your agents can verify.
Governance you can prove
Role-based operator access, organization isolation, plan retention, and usage visibility — controls a solo local node does not provide out of the box.
Evaluate flow
How organization policy checks work
Your Limetry (open source) agents already call evaluate. Limetry Cloud adds organization sign-in, operator roles, an approval inbox, and plan-based audit retention — so teams get allow, deny, or approval_required with a trail they can defend.
Action intent in
1. Agent tool call
Any SDK, MCP, or adapter loop
Your agent or adapter submits an action intent before an irreversible tool runs. Send only fields the policy needs — no secrets, tokens, or chat transcripts in metadata.
AI call
Evaluate
Policy
Audit
Enforce
What evaluate actually enforces
Limetry does not custody funds or co-sign payments. evaluateAction is a pre-flight check: allow, deny, or approval-required, plus a signed receipt. Limetry Cloud adds organization isolation, operator roles, and plan retention (14 / 90 / 365 days). Agents still enforce deny. Audit stores a privacy-safe record — policy matching sees the full intent; storage does not keep open metadata bags.
Use cases
Allow, deny, and wait in the loops teams already run. Limetry Cloud evaluates the same action intents — start on Limetry (open source) and move operators here when you need organization sign-in, roles, and retention you can defend.
Agent surface: Copilot coding agent
allow
PR typecheck
Copilot calls run_ci_privilege. Limetry allows ci_privilege so tests run. Deploy secrets stay unavailable.
deny
Fork production deploy
request_production_deploy on pull_request is untrusted. The job fast-fails and never sees production credentials.
approval required
Main production deploy
Trusted push to main still returns approval_required. The coding agent prints approval_id and exits closed.
Why teams choose Limetry Cloud
Keep the open-source engine. Add organization governance, approvals, and audit.
Adopt Limetry (open source) locally, then point the same SDK at Limetry Cloud when you need multi-team isolation, an operator approval inbox, or retention you can defend — no second product to learn.
Organization policy API
Included on Free
Point the Limetry SDK at Limetry Cloud. Scoped tokens, signed receipts, and organization isolation — with shared policies your whole team can trust.
Approval inbox
Portal + mobile
approval_required waits for an operator here, not in a chat transcript. Same action intent your open-source Limetry loop already emits.
Retention you can defend
14 / 90 / 365 days
Privacy-safe audit records purged on your plan's schedule. CSV export and organization sign-in for operators on Business.
Interactive demos
See the portal your operators will live in
Dashboard, scoped tokens, policy apply, and the approval inbox — Limetry Cloud surfaces on the same action intent your Limetry SDK already speaks. Adopt Limetry (open source) locally, then point the same SDK at Limetry Cloud when you need organization roles, approvals, and retention you can defend.
1. SDK evaluate
2. CLI & Cloud API
3. MCP / GPT Actions
4. Approval inbox
5. Portal policies
Replay
agent.js — zsh
# Install the SDK
Limetry
D
T
R
P
Audit
18
Evaluated
3
Denied
ACTIVE POLICY
allow http_get · deny deploy
audit_mode: minimal
RECENT
http_get
ALLOW
shopify.refund
ALLOW
Evaluate before the tool runs
Call createRemoteEngine from any agent loop. The server requires a UUID intent_id and issued_at.
Quick start
1. GET AN API TOKEN
Limetry Cloud dashboard → Tokens → policy:evaluate
2. APPLY A POLICY
limetry policy apply --allow http_get --deny deploy --block-resource 'github.com/acme/*'
3. EVALUATE AN INTENT
echo '{"policy_id":"PASTE_ID","action_type":"deploy","resource":"github.com/acme/api@sha"}' | limetry eval
4. TAIL AUDIT
limetry audit tail --event-type policy.evaluated
Security boundary (plain English)
Policy matching sees the full action intent. Audit stores a privacy-safe record (audit_mode=minimal by default). Outcomes include a signed receipt so middleware can detect a tampered allow. Limetry Cloud adds organization isolation, operator roles, and plan retention — it does not hold private keys or co-sign payments. Your agent or CI/CD job must still enforce deny.
