Changelog

What we shipped, in plain English.

Jul 2026

Managed AI — Ask an Expert, Team Briefing & Cross-Check analysis modes

productmanaged-aipersonas
  • Platform-funded managed AI inference is now included on every tier — no provider API key required to start using expert personas. Starter: 100 credits/mo · Professional: 300 credits/mo · Enterprise: 1,000 credits/mo.
  • Ask an Expert (1 credit) — route a question to a single expert persona, grounded in its document library. All queries are scanned before forwarding, raw content discarded after response.
  • Team Briefing (N credits, one per persona) — fan a question out to multiple expert personas simultaneously, each answering from its own domain. Results consolidated in a single briefing.
  • Cross-Check (3 credits) — send a question to two independent AI models and produce an agreement-scored synthesis. Surfaces meaningful disagreements so you know when human review is warranted.
  • Analysis modes available at /app for all published personas. Every query runs through mandatory DLP scanning and zero raw content is retained after the response completes.
Jul 2026

Tenant dashboard, use-case lanes & AI Personas portal

productdashboardpersonas
  • Dashboard restructured around four focused tabs: Keys (provider key management with use-case lane assignment), Personas (org AI persona builder), Endpoint (tenant gateway URL + service account API keys), and Logs. Advanced features (Gateway workflow view, Round Table multi-persona sessions) preserved under a collapsed sidebar section.
  • Three use-case lanes per ingested provider key: Security Scan (DLP enforce + block, default), Passthrough Monitor (every call scanned and recorded, but no violations blocked — for observe-first rollouts), and Persona (routes calls through a published /app persona). Keys can serve multiple lanes simultaneously with independent TPM budget caps per lane.
  • Org admins can publish AI personas to team members at /app on every tier. Every conversation is scanned before forwarding — team members get a governed AI interface, not a raw provider API key. Persona limits: Starter — 1 persona, Professional — up to 3 personas, Enterprise — up to 10 personas.
  • Tenant gateway endpoint URL and service account API keys are now surfaced in the Endpoint tab with a copy-ready quickstart (curl + Python) pre-filled with the org's own gateway URL.
Jun 2026

Anthropic (Claude) restored to approved provider list

securityproviders
  • Anthropic removed from the DoW supply-chain block; Claude models (claude-3-5-sonnet, claude-3-opus, etc.) are now fully supported as a direct BYOK provider alongside OpenAI, Google Gemini, Groq, and others.
  • All existing security scanning, behavioral LLM identity verification, and geopolitical enforcement continue to apply to Anthropic calls — the security layer is unchanged.
  • For classified IL4/IL5 workloads, Azure OpenAI Government or STARK/GenAI.mil remain the recommended path; Anthropic direct is approved for commercial and unclassified enterprise workloads.
  • Quickstart guide updated: Anthropic SDK now works with a direct providerType: anthropic BYOK key; AWS Bedrock path remains available as an alternative for GovCloud/FedRAMP High.
  • Anthropic AI crawlers re-allowed in robots.txt across all Blackburn Tactical properties.
May 2026

SMB brand & marketing surface refresh

brandmarketing
  • Homepage rewritten for SMB software teams. Tactical / Defense narrative moved to /defense.
  • New public surfaces: /trust, /security, /status, /changelog, /docs.
  • Compliance language hygiene pass: "FedRAMP" claims replaced with "FedRAMP-aligned architecture" / "FedRAMP Moderate readiness on roadmap"; "HIPAA compliant" replaced with "HIPAA-ready".
  • Header IA simplified: Product, Pricing, Docs, Defense, Sign in, Get started.
May 2026

Self-serve signup, public pricing & Stripe billing

billingsignup
  • Self-serve signup via Google and GitHub OAuth; unified /signup route picks the best-available channel.
  • PCI DSS-compliant subscription billing via Stripe Checkout; webhook handlers race-safe idempotent under pg_advisory_xact_lock.
  • Free-tier monthly call cap enforced server-side at the gateway with a 429 monthly_cap_exceeded envelope.
  • Tactical aliases (RECON / Asymmetric Ops / Full Spectrum) preserved as 1:1 aliases of Starter / Professional / Enterprise for backward compatibility.
Apr 2026

OpenAI & Anthropic drop-in compatibility

gatewaysdk
  • OpenAI SDK and Anthropic SDK drop-in surfaces at /v1/chat/completions, /v1/messages, and /v1/models. Streaming, tools, JSON mode, and vision supported.
  • Customers point existing SDKs at https://firemission.us/v1 by changing only the API key.