Sign inGet Control
SECURITY

Built for teams that can't ship their data to a vendor cloud

Self-hosted by design, encrypted at rest, auditable end to end — and AI agents that can propose but never deploy.

ARCHITECTURE & DATA SOVEREIGNTY

Your data never has to leave

Talon Studio ships as a single-container appliance with an embedded database and one /data volume. Self-hosted, your events, rules and credentials stay inside your network — there is no phone-home, and nothing about your environment is pooled with other customers.

One container, your infrastructure

The appliance bundles the app, an embedded Postgres and the translation sidecar in a single image with one persistent volume — deployable in your own cloud, on-prem or air-gapped.

Offline, signed licensing

Licenses are ed25519-signed and verified locally against a baked-in public key. There is no license server and no call home — the appliance runs fully disconnected.

Default outbound is public content only

The only default outbound calls pull public detection content (SigmaHQ, Atomic Red Team, CISA KEV) — and each sync loop can be disabled for air-gapped environments.

Your AI provider, your keys

AI features talk only to the provider you configure with your own API key — Anthropic, OpenAI, xAI, or a self-hosted OpenAI-compatible endpoint inside your network.

ENCRYPTION & SECRETS

Credentials are encrypted at rest and never returned

Everything sensitive the platform stores — SIEM connections, AI provider keys, threat-feed auth — is encrypted at rest with a key you control.

Encrypted at rest

SIEM, AI and integration credentials are encrypted with your ENCRYPTION_KEY before they touch the database, and are never logged or returned by any API.

Secrets stay out of Git

Feed and connection credentials never travel in the Git-backed YAML — configuration is versioned, secrets are not.

Stateless inbound tokens

The SOC alert-push webhook authenticates with a stateless per-workspace HMAC token — no inbound credential to store or rotate.

IDENTITY & ACCESS

SSO, MFA and role-based access

Access is default-deny and role-scoped at both the org and workspace level.

SSO + TOTP MFA

Sign in with GitHub, Google or Microsoft Entra, or built-in credentials with TOTP MFA (RFC 6238). Enrolment issues single-use recovery codes, stored only as hashes.

Org-wide require-MFA policy

Admins can enforce MFA for every credentials user — un-enrolled users are held at their profile page until they enrol.

RBAC everywhere

Org roles (admin / engineer / analyst / readonly) plus per-workspace roles (viewer / engineer / owner), with custom roles that can narrow who holds approve and deploy.

Scoped API tokens

Public-API tokens carry explicit scopes and can be pinned to a single workspace; admin-only scopes can only be granted by an admin.

AUDITABILITY

An append-only record of who did what

Every significant action lands in an immutable audit log — and every content change is a Git commit with an author, a diff and a history.

Append-only audit log

Actor, action, target, IP and timestamp for every significant event, retained for a configurable window (default 365 days).

Exportable for compliance

Walk the full trail through the cursor-paginated audit-log API and archive it on your own schedule.

Git as the change record

Detections live as Git-backed YAML; every edit, review and merge is permanently attributable — with one-click rollback.

AI AGENT SECURITY

Propose-only agents, by construction

Talon Claw's agents can read your evidence and draft work — but no agent, interactive or autonomous, can approve, merge or deploy. That boundary is enforced in the tool layer, not by prompt etiquette.

Propose-only tool catalog

Every write tool stages a draft change request for human review. No tool in the catalog performs an unmediated write.

Layered prompts, delimited data

An immutable system prompt sits above per-agent role instructions; operator-supplied content and external data (SIEM rows, MCP output) are handled as clearly-delimited untrusted data.

Bounded autonomy

Autonomous runs are capped by per-org concurrency and daily-run limits, with a global kill switch that stops all agent runs at once.

Guarded external tools

Tools from registered MCP servers are curated, and their action tools are queued for human approval — never auto-run.

Found a vulnerability?

We welcome responsible disclosure. Report security issues to our security team and we'll work with you on a fix and coordinated disclosure. Our security.txt has the details.

security@talonlabs.dev

Bring your own environment

Self-host Talon Studio and prove every claim on this page against your own instance — the docs cover the full security model.

Book a walkthrough Read the docs