ContactRequest access

Security & Compliance

Security and compliance at Elaichi

Certifications and security program overview.

Controls
00
Certifications
05
Subprocessors
04

01

Certifications

05 active

  • SOC 2 Type II

    ● Active

    Apr 1, 2026 → Mar 31, 2027

  • ISO 27001

    ● Active

    Jan 22, 2026 → Jan 22, 2027

  • GDPR

    ● Active

    Active

  • HIPAA

    ● Active

    Active

  • CASA Type II

    ● Active

    Active

02

Compliance program

03

Controls

No controls published yet. Controls appear here once synced from the connected GRC tool and published by an admin.

04

Security & data protection

16 practices

  • Single Sign-On (SSO)SAML SP and OIDC RP implemented in-house, so no third-party auth vendor sits in the login path. DNS-verified domains and enforced SSO per domain.
  • SCIM v2 ProvisioningUser and group sync with JIT provisioning and group-to-role mapping, so joiners and leavers are handled by your identity provider.
  • Multi-Factor AuthenticationTOTP MFA with single-use recovery codes, passkeys, and step-up authentication on sensitive actions.
  • Role-Based Access Control (RBAC)36 permissions across 8 predefined roles plus custom roles. Every call is checked against the live role of the person the agent acts for.
  • Tool-Level RestrictionsRestrictions apply to individual tools, not whole applications, at organization, role, and individual level — individual wins — and are enforced at connect, advertise, and execute time.
  • AES-256-GCM Credential VaultThird-party credentials are encrypted at rest in a dedicated vault and are never returned by the API. Eligible plans wrap the key with a customer-managed AWS KMS key (BYOK).
  • TLS in TransitAll data in transit is encrypted with TLS.
  • One-Time, Revocable TokensSession, API, MCP, and invite tokens are stored as keyed hashes and shown exactly once at creation. Endpoints can be rotated, revoking the prior token before issuing a replacement.
  • No Copies, No Cache, No IndexRecords are fetched live from the source system when an agent asks. Tool-call logging is metadata-only — arguments, payloads, prompts, and completions are never written.
  • Append-Only Audit LogEvery privileged action and every MCP tool call is recorded with actor, tool, connection, status, and duration, and can be forwarded to a destination you configure, such as Datadog.
  • Regional Data PinningAn organization picks US, EU, or APAC at creation. US and EU pin the organization's Durable Object to that Cloudflare jurisdiction; APAC is a placement hint, not a residency guarantee.
  • EU-Resident Audit StoreThe append-only audit-log store is a single European Union instance, self-hosted on OVH, serving all regions.
  • Point-in-Time RecoveryCloudflare D1 and Durable Objects carry 30 days of point-in-time recovery.
  • Bring Your Own Model KeysInference runs on your own keys across four model providers. Nothing is routed through Elaichi, so prompts and results never become our data.
  • Safe OffboardingRemoving a member runs a preflight: personal connections that shared toolboxes depend on must be transferred or deleted first, and unreferenced connections are cleaned up, vault account included.
  • Rate LimitingRate limits apply on both the control plane and the data plane.

05

Security overview

Elaichi gives an agent exactly the access the person it acts for already has — nothing more. Every call is checked against that person's live role at connect time, at tool-listing time, and at execution time, so a permission removed in the morning is gone from the agent on the next request. Nobody ever handles a credential. Connecting an application happens in a hosted sign-in flow; the credential lands in a dedicated vault encrypted at rest with AES-256-GCM and is never returned by the API — not to a user, not to an admin, not to a model. Tools run server-side with the credential resolved at the moment of the call. Session, API, MCP, and invite tokens are stored as keyed hashes and shown exactly once. Nothing is copied, cached, or indexed. Deals, tickets, and pages are read live from the source when an agent asks for them, so there is no shadow copy of your systems inside Elaichi. Tool-call logging is metadata-only: request arguments, response payloads, prompts, and completions are never written. The platform runs on Cloudflare, with organization data pinned to the EU or US Cloudflare jurisdiction at organization creation. The append-only audit-log store is a single European Union instance, self-hosted on OVH, serving all regions. Audit and tool-call metadata are retained for 90 days, and D1 and Durable Objects carry 30 days of point-in-time recovery. Our security program is independently audited — SOC 2 Type II by Prescient Assurance and ISO 27001 by Americo Quality Standards Registech — and continuously monitored through Sprinto, with each control mapped to a live evidence check rather than a point-in-time screenshot. We also hold a CASA Type II assessment. The full report and supporting documentation are released under NDA on request. We give at least 30 days' advance notice before adding or replacing any subprocessor. Vulnerability reports are accepted under a published disclosure policy at /.well-known/security.txt, and we do not pursue legal action against researchers acting in good faith.

06

Subprocessors

04 vendors

  • Cloudflare, Inc.

    Purpose
    Hosting, compute, storage, queues, email delivery, rate limiting, and access control
    Location
    Global, with EU/US jurisdiction pinning for organization data

  • OVH SAS

    Purpose
    Hosting of the self-hosted audit-log store
    Location
    European Union

  • Stripe, Inc.

    Purpose
    Subscription billing and payment processing
    Location
    United States

  • Yin Yang Technologies Private Limited

    Purpose
    Engineering, operations, and support personnel; wholly owned affiliate of Yin Yang, Inc.
    Location
    India

07

FAQ

10 questions

  • 01Does Elaichi store our SaaS credentials?

    Credentials are held in a dedicated vault, encrypted at rest with AES-256-GCM, and are never returned by the API — not to users, not to admins, not to agents. Connecting happens in a hosted sign-in flow, and tools run server-side with the credential resolved at the moment of the call, so no person and no model ever holds a key.

  • 02Does Elaichi copy or index our data?

    No. Deals, tickets, and pages are read live from the source system when an agent asks for them. There is no shadow copy of your CRM inside Elaichi, and tool-call logging is metadata-only — request arguments, response payloads, prompts, and completions are never written to the log.

  • 03How do you stop an agent from exceeding a user's permissions?

    Every call is checked against the live role of the person the agent acts for: 36 permissions across 8 predefined roles, plus any custom role you create. Restrictions work on individual tools at organization, role, and individual level, and are enforced at three points — connecting, listing tools, and running a call. A blocked tool never appears in the agent's list.

  • 04Where is our data stored?

    An organization selects a region — US, EU, or APAC — at creation, and it cannot be changed afterwards. For US and EU the organization's Durable Object is pinned to the matching Cloudflare jurisdiction; APAC is a placement hint, not a residency guarantee. The global index of organization and user records, the session store, and stored logos are not region-pinned, and the append-only audit-log store is a single European Union instance serving all regions.

  • 05How long do you keep audit and tool-call data?

    Audit and tool-call metadata are retained for 90 days, then aged out. Deleting an organization does not delete its audit history: each record ages out 90 days after it was written, and on request we can purge it sooner. Cloudflare D1 and Durable Objects carry 30 days of point-in-time recovery.

  • 06How do you manage subprocessors?

    Our subprocessors are listed publicly in the Subprocessors section and in Annex III of our DPA. We give at least 30 days' advance notice before adding or replacing one, and you may object on reasonable data-protection grounds within the notice period; if we cannot resolve the objection you may terminate the affected service without penalty and receive a pro-rata refund.

  • 07Do our prompts go through Elaichi's model account?

    No. You bring your own keys for four model providers, and nothing is routed through us for inference. Prompts and results never become our data, and there is no markup and no lock-in.

  • 08How do we report a security vulnerability?

    See https://elaichi.ai/.well-known/security.txt or email support@elaichi.ai. We do not pursue legal action against researchers acting in good faith under that policy.

  • 09Can we review your DPA and security documentation?

    Yes. The Data Processing Agreement is published at https://elaichi.ai/dpa/, and gated security documentation is released under NDA on request from support@elaichi.ai. On-site audits are not offered; section 12 of the DPA sets out how audit rights are satisfied.

  • 10Which security certifications does Elaichi hold?

    Elaichi maintains SOC 2 Type II and ISO 27001, holds a CASA Type II assessment, operates in alignment with GDPR, and is HIPAA compliant. The current status and validity for each is shown in the Certifications section above, and the compliance program behind them is listed under Compliance program. The full SOC 2 Type II report and supporting security documentation are available under NDA on request from support@elaichi.ai. We are not certified under the EU-US Data Privacy Framework.

08

Documents

No documents available.

09

Policies

23 documents

The information-security policies behind our program. Documents are shared on request — submit a request and we’ll send a secure link once approved.

  • Acceptable Use PolicyRequest
    ISMS · v2.0
  • Access Control PolicyRequest
    ISMS · v2.1
  • Application Security PolicyRequest
    Security · v1.1
  • Asset Management PolicyRequest
    ISMS · v1.3
  • Business Continuity & Disaster Recovery PlanRequest
    Operations · v1.3
  • Change Management PolicyRequest
    Operations · v1.3
  • Code of Business ConductRequest
    People · v1.4
  • Cryptography & Key Management PolicyRequest
    Security · v1.2
  • Data Backup PolicyRequest
    Operations · v1.3
  • Data Classification & Handling PolicyRequest
    Privacy · v1.2
  • Data Protection & Privacy PolicyRequest
    Privacy · v2.0
  • Data Retention & Disposal PolicyRequest
    Privacy · v1.1
  • Endpoint & Mobile Device Security PolicyRequest
    Security · v1.0
  • Human Resources Security PolicyRequest
    People · v1.3
  • Incident Response PlanRequest
    Operations · v1.4
  • Information Security PolicyRequest
    ISMS · v3.2
  • Network Security PolicyRequest
    Security · v1.3
  • New Hire Onboarding PolicyRequest
    People · v1.1
  • Risk Assessment & Management PolicyRequest
    ISMS · v1.4
  • Secure Software Development PolicyRequest
    Security · v1.4
  • Security & Privacy Awareness PolicyRequest
    People · v1.2
  • Vendor & Subprocessor Management PolicyRequest
    ISMS · v1.2
  • Vulnerability & Patch Management PolicyRequest
    Security · v1.1

10

Legal & privacy

08 documents