Pomerado launches MCP that turns websites into agent APIs
The YC F26 startup says hosted, self-healing integrations beat browser agents on speed and cost, with sign-in and 2FA handled for you.
Stories per day · 30 days
22 RSS
Relatedai-agents19prompt-injection14agentic-ai6llm-agents6shadow-ai6agents5model-context-protocol5open-source5agent-security4ai-security4claude4codex4supply-chain4ai-governance3anthropic3coding-agents3command-injection3cryptomining3
217 stories, newest first · one card per story
Stories tagged #mcp
Unpinned MCP server launches can expose API keys, Corelight finds
Of 35 servers reviewed, 23 recommended installs pulled fresh registry dependencies; pinning the main package did not lock the rest.
SOCRadar promotes an MCP server and a new MSSP MCP server
The captured page is mostly site navigation, so the 126-tool claim is not described in the body.
Atuin Threads syncs coding-agent sessions across machines
From Atuin 18.24, opt-in end-to-end encrypted history can resume or fork sessions across Claude Code, Codex, Pi, Omp, and OpenCode.
1
Bay raises $50 million to control enterprise AI agents
Founded in February 2026 by three Unit 8200 alumni, Bay inventories agents and can deny actions in milliseconds.
Orca adds discovery for shadow AI apps and agents
Agentless scanning looks for builder fingerprints and maps agent permissions, credentials and access to sensitive cloud data.
Obsidian launches runtime policy engine for AI agent actions
Policies can allow, block, or require approval, including for OpenAI Codex and MCP tool calls.
Exploit-DB lists RCE in Ruflo MCP Bridge 3.16.2
The public entry labels version 3.16.2 remotely exploitable but gives no CVE, fix, or confirmed attacks.
Resend CEO explains why coding agents recommend its email API
In one study Claude Code ranked Resend first in 64 of 102 extractable email recommendations.
1
Adobe previews local MCP agents and always-on Firefly automation
OpenAI, Google, and Anthropic agents could edit local Creative Cloud files; a private beta is set for early 2027.
1
Medium launches TK, an AI assistant inside the writing editor
Users type @TK for lookups, editorial feedback and style checks; the free web app can note AI use and connect an agent via MCP.
1
Mittwald says MCP flaw let an attacker hit 56 customer projects
The German host says the intruder was undetected for eight days; AI use is inferred from log speed only.
Paper tests Agent Detector for unsanctioned LLM agents on Windows
On a small controlled bench it caught every synthetic agent in most runs while two behaviour-only rules flagged every twin.
0sec Labs open-sources 0, an AI security harness
The MIT or Apache-2.0 research preview uses a bring-your-own model and self-reports 93 of 95 XBOW solves.
ShadowBench-MCP measures shadow invocations in MCP-style tool routing
On a Groq-hosted qwen3-32b router, efficiency-first routing had a 29.2% shadow-invocation rate versus 4.2% when policy-aware.
PRISM paper cuts LLM agent tool privileges using causal roles
On AgentDojo, tau-bench and ToolEmu, authors report 78.3% fewer tools per role and a 15.9% blast radius.
Paper builds a 24-case dataset of MCP security vulnerabilities
Authors say 87.5 percent are Critical or High, and injection plus path traversal account for 66.7 percent.
Review maps 284 attacks on agentic AI systems
Tool, MCP, and action-layer attacks dominate, while only 278 of 660 attack-safeguard links are directly assessed.
Sysdig AI Defense adds runtime guardrails for AI agents
It can allow, hold, or block an agent action before it finishes and keeps a kernel record the agent cannot change.
Postman desktop app installs plugins into AI coding agents without permission
Postman’s installer auto-approves its own startup hook for Codex, raising consent concerns for developers.
1
AI coding agents install unpatched deprecated MCP servers
Glow Labs research found AI agents installing deprecated MCP servers in 98% of sessions, with four high-severity vulnerabilities including RCE, downloaded over 17.3 million times, no patches or exploitation seen.
MCP Tool Poisoning Enables Attacker Tool Calls in Agentic AI
Research finds 22.9% attack success rate with 2,113 vulnerable cells across 12 frameworks and 45 scenarios.
PentestPad pitches self-hosted AI so client findings stay in-house
The reporting platform can run air-gapped, point its assistant at a local model, and expose findings over MCP.
Three-Tier Framework Maps MCP Vulnerabilities
MCP's blended data/control planes create unmitigable prompt injection and confused deputy issues beyond server fixes.