Changelog

Every feature, fix, and improvement. Shipped fast, documented properly.

v3.8.0 Latest
March 29, 2026
Strat 4 — Browser Primitives, Memory 2.0, Marketplace, Safety Layer

Added

  • Browser/Desktop primitives: Swarm Visualizer SSE stream, MCP Desktop bootstrap (slop mcp bootstrap --ide=claude-desktop), resource devotion for GPU/CPU sharing
  • Memory 2.0: GraphRAG graph queries (POST /v1/memory/graph-query), auto-summarize namespaces, snapshot and pin for permanent storage
  • Grok-specific features: POST /v1/grok/optimize, POST /v1/grok/critique, and POST /v1/grok/overlord for autonomous multi-LLM orchestration
  • Safety and guardrails: sandbox execution with vm.createContext isolation, circuit breakers, reputation slashing for underperforming agents, chaos testing for army deployments
  • Marketplace: template browsing and invocation, bounty board for agent work, Plugin Forge for building and publishing custom tools
  • Agent templates library: research swarm, content machine, security audit, data pipeline — pre-built agent configurations with custom template creation
  • 18 new endpoints bringing total to 1,303

Improved

  • Endpoint count updated across all documentation and localized files
  • MCP server now supports bootstrap mode for automated IDE configuration
  • Memory tier extended with graph, summarize, and snapshot capabilities (still free, 0 credits)
  • Multi-LLM orchestration now includes Grok-driven autonomous mode
v3.7.0
March 29, 2026
Engine: Real — Every Feature Backed by Actual Computation

Added

  • 4 live LLM providers: Claude Opus 4.6, GPT-4o, Grok-3, DeepSeek — all verified on production
  • Army survey with real LLM inference — each persona gets individual reasoning via POST /v1/army/survey
  • Army quick-poll with LLM agent voting — agents reason about options before casting votes
  • Knowledge graph POST support — POST /v1/knowledge/query with free-text search
  • Fine-tuning actually submits to OpenAI API — uploads JSONL, creates job, returns provider_job_id
  • Cost optimizer live benchmarking — pass benchmark:true to test providers in real-time
  • Smart router queries real audit_log data — latency, success rate, cost per provider from last 7 days
  • POST /v1/random endpoint (was GET-only)

Fixed

  • Grok dedicated endpoint: fixed env var mismatch (XAI_API_KEY now checked)
  • 9 CLI bugs: balance JSON mode, call arg parsing, hive todo writing, spinner, verbose flag, extractMeta
  • Compliance endpoints: SOC2/HIPAA now perform real runtime checks instead of hardcoded passed:true
  • Confidence scoring: simulated endpoints no longer get 99% confidence (now 50%)
  • Removed fabricated uptime claim (99.97%), boot "0 mocks" message, case study fake numbers
  • Compare endpoint: no longer ranks LLM quality by response length

Improved

  • Pricing aligned across all 26 files: 1K/$9, 10K/$49, 100K/$299, 1M/$1,999
  • Handler counts corrected: 925 compute handlers, 1,303 total endpoints
  • 17 localized llms files updated (1248→1269, 927→925)
  • _engine field added to wallet, knowledge, and random endpoints
  • Python SDK: added memory_list() and categories() for Node parity
  • 9 missing commands added to interactive CLI REPL
v3.2.0
March 27, 2026
Series B Engineering

Added

  • Comprehensive test suite with full endpoint coverage and edge case testing
  • Node.js and Python SDKs with typed interfaces and automatic retries
  • OpenAPI 3.1 specification at /v1/openapi.json
  • Structured JSON logging with request tracing and correlation IDs
  • Enterprise landing page, changelog, roadmap, and status page
  • Security audit tooling and penetration testing framework

Improved

  • Input validation hardened with schema enforcement on all endpoints
  • Response headers now include rate limit, request ID, and timing metadata
  • Error responses standardized across all tools across 82 categories
  • Documentation coverage expanded with integration guides
v3.1.0
March 2026
Security Hardening & Platform Expansion

Added

  • Expanded to 82 categories of tools
  • Credit system for earning and spending credits
  • Copilot mode for interactive agent assistance
  • PII detection and redaction APIs
  • RAG primitives for retrieval-augmented generation
  • SVG chart generation endpoints

Improved

  • Security hardening: CSP headers, timing-safe auth, HTTPS enforcement
  • Memory isolation between API keys
  • Rate limiting with per-tier burst allowances
  • 925 compute handlers benchmarked for performance
v3.0.0 Major
February 2026
Major Expansion

Added

  • Hackathon features: hive workspaces, agent teams, prediction markets
  • Competitor gap analysis and feature parity coverage
  • RapidAPI marketplace integration
  • Batch and pipe workflow chaining endpoints
  • Agent templates and autonomous task execution
  • 10K compute army for parallel verified agents

Improved

  • API catalog expanded significantly with new tool categories
  • MCP server integration for Claude Code
  • Credit system refined with better tier pricing
v2.0.0 Major
January 2026
Initial Launch

Added

  • 672 APIs across initial tool categories
  • Persistent agent memory with SQLite backing (85 tables)
  • Agent mode with autonomous task execution
  • Credit-based pricing with 500 free credits on signup
  • Self-hostable deployment with zero external dependencies
  • CLI tool and MCP server for IDE integration
  • Zapier integration endpoints