Anchor Action Index: replay proven web workflows as code via MCP
At a glance
Anchor (Oct 15) ships Action Index — shared execution memory for computer-use agents. Successful site workflows are stored as parameterized actions and replayed as deterministic browser code, with the model only intervening when a page changes. On Online-Mind2Web, gpt-5.6-luna + Action Index hits 93.3% success at $0.016/task vs claude-opus-5 at $1.57; paired Skyvern runs cut p90 tokens ~74%. Free tier included; MCP tools search_action_index / get_action_index / execute_action_index at api.anchorbrowser.io/v1/mcp. This workflow is listed under web-agents, MCP, browser-infra, cost-reduction, action-memory.
GitHub BenniAlencar/agent-cyber-sentinel (Apache-2.0): Tier-0 runtime defense for autonomous agents and MCP servers—prompt-injection/jailbreak firewall, secret leak interceptor with Shannon-entropy redaction, SSRF metadata blocking, tool-call sanitization, and tamper-evident Proof-of-Defense (POD-9) envelopes. Framed as the missing security layer when agents take real actions; CI + pytest suite in-repo.
Self Labs’ Agent Vault gives agents mock tokens for checkout and sign-in; real card numbers and passwords never reach the model. Caps, merchant allowlists, and per-purchase approval. Connect Grok Bot via MCP or Meta Muse via custom connector in ~2 minutes (also Claude Code, Hermes, Codex).
Lumovi (OSS Apache-2.0, v1.22.1) is a calm K8s dashboard that doubles as an MCP server: assistants read your clusters as you, and every change waits for approval with a diff, reason and equivalent kubectl. Namespace/RBAC rules constrain what agents may touch; desktop one-click MCP setup or Helm-in-cluster for teams. Pattern: ops UI + HITL gate for agentic cluster work.