Released October 2026. Cited answers.
What's new:
- Query graph — a planner turns one query into a validated graph, executes the nodes, and returns an evidence packet where each claim carries the source_ids it rests on. Every claim can be traced back to the page behind it.
- Dry-run planning — POST /v1/graph/plan validates a graph and returns a cost estimate without touching the network, so a run can be sized before it is paid for.
- Gap reporting — a branch that timed out or ran dry records the node, the query and the reason instead of quietly returning less evidence. A value the engine cannot derive is null, never a plausible-looking guess.
- session_handoff — packages a finished run into a brief another agent can act on: resolvable evidence ids, open questions, contradictions, remaining budget, and the next suggested node.
- Graph everywhere — the Cloud API proxies it at /v1/graph/plan (1 credit) and /v1/graph/search (10 credits), and the Python, JavaScript and Go SDKs expose it as plan_search/search_graph, planSearch/searchGraph, and PlanSearch/GraphSearch.
- 22 MCP tools self-hosted, 18 on Cloud — the Cloud proxy serves the subset it has routes for, so nothing is advertised that cannot run.
Released October 2026. Truth, metering, and reach.
What's new:
- Credit enforcement — spend-before-serve metering with automatic refunds on every search, batch, stream, social, and MCP call. Empty balances get a real 402 INSUFFICIENT_CREDITS.
- 16 MCP tools — batch search, social search, agent research, usage, history, engines, status, account, webhooks, and monitors, plus docs resources and task prompts.
- Observability — structured logs, error tracking, persisted engine health, real backend probes, and a status page with incident history, RSS, and email alerts.
- Docs platform — full-text search, OpenAPI 3.1 generated from validators, and a live API reference with copy-paste snippets.
- Release gates — Playwright e2e, accessibility checks, dependency audit, and bundle budgets green in CI.
Released September 2026. The complete platform release with production-ready developer access.
What's new:
- Premium CLI authentication — browser-based device-code flow, realtime Firebase session sync, resilient loading states, and a fresh API key on every CLI authorization.
- Unified packages — the CLI, JavaScript SDK, Python SDK, and Cloud API all publish as 0.3.1.
- Realtime auth infrastructure — Firebase ID-token refresh keeps browser sessions and CLI handoffs synchronized.
- Reliable local development — Firestore emulator support works end-to-end for device code, authorization, and token polling.
- Release-hardened API — synchronized credit pricing, stronger WAF normalization, internal auth hardening, and correct device-key issuance.
- Refreshed product surface — unified Jiro branding, favicon, CLI logo, responsive dashboard navigation, and current documentation.
Released September 2026. Focused on polish, documentation, and brand identity.
What's new:
- Landing page redesign — premium UI with animated gradient hero section and agent logos showcase. New CTA flows for waitlist and self-hosted install.
- Docs overhaul — corrected SDK class names (JiroClient for Python/JS, NewClient for Go), updated all code examples to 2026 dates, added missing error handling sections.
- Build fixes — resolved sitemap generation timeout by switching from dynamic fetch to fs.readFileSync, fixed loader component for static content delivery.
- Brand logos — added real SVG logos for all 9 search engines and 12 social platforms. Replaced placeholder images with high-fidelity brand assets.
- Minor fixes — fixed typo in rate-limits table, corrected credit cost for Google search (10 credits), added missing request_id to error response examples.
Released August 2026. Major landing page and infrastructure overhaul.
What's new: - Full landing page redesign — 15 marketing components rebuilt from scratch with modern design system. Consistent typography, spacing, and colour palette across all sections. - Data-driven features — YAML-based content system for easy updates without code changes. Features, pricing, and testimonials are now driven by structured data files. - Animated counters — real-time GitHub stars, forks, and contributor counts displayed with smooth counter animations on the landing page. - Vercel deployment — migrated to Vercel for edge-first hosting. searchjiro.vercel.app is now the primary domain with automatic preview deployments for PRs. - Performance — reduced initial page load by 40% through code splitting and lazy loading of below-the-fold sections.
Released July 2026. Windows compatibility, stealth improvements, and new CLI features.
What's new:
- Windows Unicode fix — added _safe_print() utility that handles cp1252 encoding on Windows terminals. Prevents UnicodeEncodeError when printing non-ASCII characters from search results.
- Stealth engine — integrated curl-cffi with browser impersonation profiles (Chrome, Firefox, Safari). Requests now mimic real browser TLS fingerprints to avoid detection.
- WebSocket streaming — search results can now be streamed in real-time via WebSocket connections. Reduces time-to-first-result for large result sets.
- Parallel search — fan-out search across multiple engines simultaneously. Results are merged and deduplicated. Enabled with --parallel flag.
- Interactive mode — guided search wizard that walks users through engine selection, query refinement, and result filtering. Run with jiro search --interactive.
- Benchmark tool — new jiro bench command for performance testing. Measures throughput, latency percentiles, and engine response times.
Released June 2026. Enterprise features, agent framework extras, and 28 tools.
What's new:
- Agent Framework Extras — pip install extras for OpenAI, Anthropic, LangChain, and LlamaIndex. Install with pip install jirosearch[openai,anthropic,langchain,llamaindex].
- Pro Deep Lock — removed boolean feature flags from config. Features are now mapped to tiers (free, pro, enterprise) with automatic feature gating.
- Server-side License Verification — HMAC-SHA256 license token middleware validates enterprise licenses on every request. Tokens are signed with a server-side secret.
- Anti-tamper Check — package integrity verification at startup. Detects modified source files and prevents running tampered installations.
- MCP Setup Wizard — jiro mcp setup provides one-command configuration for 8+ MCP clients including Claude Desktop, Cursor, and Continue.dev.
- jiro doctor — built-in diagnostics command that checks Python version, dependencies, Redis connection, configuration validity, and package integrity.
- 9 Search Engines — Bing, Google, DuckDuckGo, Brave, YouTube, Amazon, eBay, Yandex, Baidu. Each with circuit breaker and rate limiting.
- 12 Social Platforms — Reddit, Twitter/X, YouTube, LinkedIn, TikTok, Instagram, Facebook, Threads, Hacker News, Bluesky, Telegram, Pinterest.
- 16 MCP Tools — 9 free tools (search, scrape, status, etc.) and 7 enterprise-only tools (deep research, social scrape, batch processing).
Released May 2026. Enterprise tier, free tier expansion, and security hardening.
What's new:
- Enterprise Tier Deep Lock — HMAC-SHA256 license validation for enterprise features. Licenses are bound to installation fingerprints and validated server-side.
- Free Tier Power-Up — expanded free tier to 100 RPM, 10,000 RPD, and 12 free features. Previously limited to 30 RPM and 5 features.
- Anonymous Access — expanded anonymous (no API key) access from 3 to 12 free features. Includes search, scrape, status, and basic AI endpoints.
- Security Headers — added Content Security Policy (CSP), HSTS, X-Frame-Options, X-Content-Type-Options, and Referrer-Policy headers to all responses.
- Self-Hosting — official self-hosting support with pip install jirosearch && jiro serve. Includes Docker image, docker-compose template, and systemd service file.
- Configuration — YAML-based configuration at ~/.jiro/config.yaml with environment variable overrides for all settings.
- Engine improvements — added circuit breakers with automatic recovery, per-engine rate limiting, and request deduplication.