Daily AI Implementation Scout Council

Today's ranked top 20

#1

newtest firstBoth runtimes30 / 35

GitHub MCP Server v2.0.1 repo

GitHub's own MCP server crossed a major version to 2.0, replacing loosely typed JSON tool outputs with typed tool schemas and typed outputs across repos, search, security alerts, comments, and discussions. The prior release, 1.14.0 on October 2, 2026, added governance tools to read and manage repository rulesets and custom properties at the repository, organization, and enterprise level.

What it does for you: Gives Axion's GitHub facing agent calls structured, validated outputs instead of loosely typed JSON, which should cut parsing errors in any skill that reads GitHub data. The new ruleset and custom property tools let an agent check repo governance settings directly instead of YY checking by hand.

In practice: Reads like a vendor quietly hardening a tool everyone already depends on, not a flashy relaunch.

For: Both runtimes. A standalone MCP server process, usable by any MCP client whether the caller is an Axion or a Hermes agent.

Security4

Quality5

Auditability4

Useful to you4

Useful to community5

Buildable now4

Hermes4

Verdict: test first. Official, highly credible vendor and a real improvement, but a major version jump from 1.x to 2.0 changes output shapes, so this earns a test pass against Axion's existing GitHub MCP usage before it goes into the daily path.

Build #1 GitHub MCP Server v2.0.1: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/github/github-mcp-server/releases/tag/v2.0.1. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Version 2.0.1 is the current latest release. The prior governance release, 1.14.0, shipped October 2, 2026. Pull request numbers in the typed schema rollout run into the 3400s, for example pull request 3407.

#2

newtest firstBoth runtimes27 / 35

mcp-warden repo

A CI first MCP supply chain integrity gate plus runtime tool result inspection. It pins an MCP server's tool surface, fails CI on drift, classifying each mutation such as a required field dropped, an enum widened, a type broadened, or a constraint relaxed, and blocks poisoned tool results such as ANSI escapes, secret echoes, or exfiltration patterns at runtime, using RFC 8785 plus SHA 256 canonicalization and SARIF upload to code scanning.

What it does for you: Axion runs several of its own MCP servers, including context-mode and team-crm. This drops into a CI pipeline as a gate that fails the build if one of those servers' tool definitions silently changed, and inspects live tool results for the exact poisoning patterns security researchers have been finding across MCP servers this year.

In practice: Small, single purpose, and precisely scoped. It reads like a security tool written by someone who has actually had to debug a changed tool schema, not a general platform pitch.

For: Both runtimes. A Python CLI plus GitHub Action that pins and checks an MCP server's own tool surface. It runs in CI or locally, with no Claude Code specific hook, so it fits Axion or Hermes equally.

Security4

Quality3

Auditability4

Useful to you5

Useful to community3

Buildable now4

Hermes4

Verdict: test first. Directly answers a documented, current risk to Axion's own MCP servers with an MIT licensed, well structured tool, but it has only 2 stars and no independent track record yet, so it earns a pilot run against one Axion MCP server before wider rollout.

Build #2 mcp-warden: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/DataScience-EngineeringExperts/mcp-warden. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · 2 GitHub stars, last updated October 4, 2026, MIT license, requires Python 3.11 or newer, ships a working GitHub Action.

#3

newtest firstBoth runtimes26 / 35

DeepTeam v1.0.9 repo

Confident AI's LLM and agent red teaming framework, already tracked in this registry at test first since August 11, 2026, reached its first stable release at v1.0.9, after an earlier 0.1.9 line that added an agentic red teaming CLI tool. It detects a wide range of vulnerabilities using research backed attack methods and ships production ready guardrails for real time input and output filtering.

What it does for you: Crossing 1.0 is the kind of milestone that should move a tool from trial to trusted. It directly covers the automated adversarial testing YY's own Claude skills and MCP servers currently get none of, the same gap the Enkrypt AI research flagged this week.

In practice: Feels like a project that spent its pre 1.0 time earning the label, not rushing to it.

For: Both runtimes. A Python red teaming library and CLI with no Claude Code specific dependency, so it can test any agent or skill, in Axion or Hermes.

Security4

Quality4

Auditability3

Useful to you4

Useful to community4

Buildable now4

Hermes3

Verdict: test first. The 1.0 milestone earns a step up from the watch verdict this registry has carried since August, but a brand new stability claim deserves one real red team run against an Axion skill before trusting the stable label fully.

Build #3 DeepTeam v1.0.9: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/confident-ai/deepteam/releases/tag/v1.0.9. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · v1.0.9 is tagged as the First Stable Release. This registry already tracked the project at test first since 2026-08-11, most recently seen 2026-10-06.

#4

newtest firstBoth runtimes24 / 35

bdg, Browser Debugger CLI repo

A Node CLI that lets Claude Code and other coding agents drive and debug Chrome directly from the shell: navigation, DOM and shadow DOM interaction, network and HAR capture, console messages, accessibility tree queries, performance and CPU or heap profiling, and raw CDP commands, all without screenshots and without an MCP server in the loop.

What it does for you: A lighter weight path for the render verify workflow YY already requires on every visual deliverable. Instead of a full headless Chrome screenshot round trip, the agent can query layout, contrast, network requests, and console errors as short shell commands.

In practice: Reads like a tool built by someone who was tired of screenshotting to debug CSS, not a wrapper pitched as a platform.

For: Both runtimes. A Node CLI that talks to Chrome over the DevTools protocol directly. It has no Claude Code specific binding, so it fits any coding agent or harness, Axion or Hermes.

Security3

Quality3

Auditability4

Useful to you4

Useful to community3

Buildable now4

Hermes3

Verdict: test first. A real fit for an existing, mandatory workflow step. Worth a trial run against one render verify pass before relying on it over the current headless Chrome script.

Build #4 bdg, Browser Debugger CLI: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/szymdzum/browser-debugger-cli. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Submitted to awesome-claude-code on October 6, 2026, and published on npm as browser-debugger-cli.

#5

newtest firstBoth runtimes23 / 35

prismor repo

A self hosted runtime control plane for AI agents that observes, or lets a human approve or block, risky tool calls before they execute, covering secret leaks, prompt injection, and supply chain issues in a local dashboard. It is agent agnostic across Claude, Codex, LangChain, and others.

What it does for you: A direct answer to the gap the Enkrypt AI research flagged this week: 73 percent of 268,210 scanned agent tools across 25,264 MCP servers carried some vulnerability. This gives a human in the loop checkpoint before a risky tool call runs, across every agent YY runs, not only Claude Code.

In practice: Reads as a genuine safety net rather than a dashboard for dashboard's sake, the kind of tool built right after someone got burned.

For: Both runtimes. A self hosted, agent agnostic control plane by its own description, built to sit in front of Claude, Codex, or LangChain agents alike, so it fits Axion or Hermes without modification.

Security4

Quality2

Auditability3

Useful to you4

Useful to community3

Buildable now3

Hermes4

Verdict: test first. The idea maps directly onto a real, currently unmitigated risk across YY's agent fleet, and the agent agnostic design fits a two runtime setup like Axion and Hermes, but it is one day old with no independent review yet.

Build #5 prismor: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/PrismorSec/prismor. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Submitted to awesome-claude-code on October 7, 2026, and describes coverage across 3 or more agent runtimes by its own README: Claude, Codex, and LangChain.

#6

newtest firstClaude (Axion)23 / 35

Claude Code mods, You should know built in mod and the --marketplace install flag plugin

Claude Code shipped mods in version 2.1.288 on October 2, 2026, with a first built in one, You should know, a side agent that watches a session and flags things the main agent or YY might miss, enabled per session with the command /plugin enable cc-plugin-you-should-know@builtin. Version 2.1.292 on October 6, 2026 added a --marketplace flag to claude plugin install, so one command adds the marketplace and installs the plugin under the same policy checks.

What it does for you: A free, official second pair of eyes inside every Claude Code session with no separate subscription, plus a one line install path for any plugin instead of two separate commands, useful across every one of YY's Claude Code sessions.

In practice: Feels like Anthropic is slowly turning Claude Code into a platform with its own extension ecosystem, one small piece at a time.

For: Claude (Axion). Mods and the plugin marketplace flag are Claude Code specific features with no Hermes equivalent.

Security2

Quality5

Auditability3

Useful to you4

Useful to community4

Buildable now4

Hermes1

Verdict: test first. A real, official feature, but mods run unsandboxed with full machine access per Anthropic's own documentation, the same basis this skill used to downgrade similar mods previously. It earns a trial on one low stakes session before any wider use.

Install: run /plugin enable cc-plugin-you-should-know@builtin inside Claude Code, and use the new --marketplace flag with claude plugin install for any future plugin. Review before enabling. Copy

Source · Mods shipped in Claude Code 2.1.288 on October 2, 2026. The --marketplace install flag shipped in 2.1.292 on October 6, 2026.

#7

newtest firstBoth runtimes22 / 35

Executor skill

An initiative scoped workflow that carries one body of work from intake through architecture, spec, planning, execution, review, and verification. State transitions are enforced by POSIX bash scripts rather than prose: dispatch is one locked script act, a worker's result enters run state only through a gate that refuses without a clean verdict, and a graph check fails when any verb loses its actuator or a reference stops resolving.

What it does for you: A stricter, script enforced version of the two level hierarchy and verify before done habits YY already runs by hand. The hard gate that refuses a result without a clean verdict is exactly the kind of mechanical check YY's own rules ask for everywhere else.

In practice: Reads like someone encoded the discipline of a good code review process into bash gates instead of hoping an agent remembers to be careful.

For: Both runtimes. Works with Claude Code, Codex, Cursor, and omp by its own description, with state enforced by POSIX bash scripts, not a Claude specific binding.

Security3

Quality2

Auditability4

Useful to you4

Useful to community3

Buildable now3

Hermes3

Verdict: test first. The gate mechanism matches YY's own verify before done philosophy closely, but it is a brand new skill with no independent usage history, worth a trial on one small initiative before trusting it on real work.

Build #7 Executor: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/Atri10/executor. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Submitted to awesome-claude-code on October 5, 2026, and merged October 7, 2026.

#8

newwatchBoth runtimes22 / 35

mcp-pin plugin

Tool integrity pinning for MCP: pins the exact tool schema an agent approved and blocks silent changes, the rug pull or tool poisoning pattern, plus a public, hash linked, signed history of third party MCP servers' past tool definitions. It answers a different question than mcp-warden, which covers servers you build yourself. mcp-pin covers servers you merely install.

What it does for you: Covers the third party MCP servers YY installs from elsewhere, complementing mcp-warden which better fits Axion's own MCP servers. Together they are two independent answers to the same Enkrypt AI finding named in this issue's dissent.

In practice: A clever, narrow idea, a public ledger for other people's tool definitions, that only works if enough people feed it.

For: Both runtimes. Ships a Claude Code plugin and a Cursor plugin from the same codebase, plus a plain CLI and GitHub Action, so it works outside Claude Code too.

Security4

Quality2

Auditability3

Useful to you4

Useful to community3

Buildable now3

Hermes3

Verdict: watch. A good idea, but too young at 3 stars with no independent validation yet to trust its signed history claims. Revisit after it gets more eyes.

Install: clone GautamTalksDev/mcp-pin and install the bundled plugin from its .claude-plugin or .cursor-plugin folder. Review before enabling. Copy

Source · 3 GitHub stars, submitted to awesome-claude-code on October 6, 2026.

#9

newtest firstBoth runtimes22 / 35

oh-my-graph repo

A Go CLI where each node of a DAG you author in YAML, or that the tool auto plans from a goal, runs as a real subprocess of your own logged in Claude or Codex CLI rather than a metered API key. API key variables are deleted from child process environments so the CLI falls back to the saved login, and run ledgers record each node's verdict and cost. Latest tagged release is v0.16.0.

What it does for you: A way to run many parallel Claude Code or Codex nodes against YY's own logged in CLI sessions instead of a separate metered API key, directly relevant to how many parallel sessions YY already runs across workspaces.

In practice: Feels like a graph runner built by someone who pays for Claude Code the normal way and did not want a second API bill on top of it.

For: Both runtimes. A Go CLI that runs DAG nodes as subprocesses of your own logged in Claude or Codex CLI, with no deeper Claude Code binding.

Security3

Quality3

Auditability3

Useful to you4

Useful to community3

Buildable now3

Hermes3

Verdict: test first. A direct fit for YY's existing multi session habit and a genuinely different cost model than a metered API key, worth a small trial graph before trusting it on a real workload.

Build #9 oh-my-graph: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/jitokim/oh-my-graph. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Latest tagged release is v0.16.0, with pull requests numbered up to 361 showing active development.

#10

newwatchBoth runtimes21 / 35

AgentHub repo

A privacy first session control plane that lists the Claude Code and Codex sessions already running on your machines by reading their own files, lets you share a short status summary of a session with paired machines, and lets agents leave notes for each other through four MCP tools or an ah CLI command, with an option to wake a sleeping Codex session when a note arrives.

What it does for you: Addresses the exact problem YY lives with: running many parallel Claude Code sessions across workspaces and needing to know their status or leave them a note without exposing prompts or transcripts to other machines.

In practice: A thoughtfully scoped answer to a real multi session headache, not a general purpose dashboard pretending to solve it.

For: Both runtimes. Explicitly cross provider by its own description, covering both Claude Code and Codex sessions.

Security3

Quality2

Auditability3

Useful to you4

Useful to community3

Buildable now3

Hermes3

Verdict: watch. The idea maps well onto YY's actual multi session workflow, but it is one day old in this registry with no independent usage yet to confirm the privacy claims hold up.

Build #10 AgentHub: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/SheldonChangL/agenthub. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Submitted to awesome-claude-code on October 6, 2026, and covers 2 providers, Claude Code and Codex.

#11

newwatchBoth runtimes21 / 35

LlamaIndex Extract v2.5 repo

LlamaIndex released Extract v2.5, its next generation document extraction agent product, alongside ExtractBench, an open benchmark scoring 14 extraction systems across 370 enterprise documents, long filings, scans, and handwriting, on completeness, accuracy, grounding, and cost per page.

What it does for you: A benchmarked way to pick a document extraction approach instead of guessing, relevant if any future Axion build needs to pull structured data out of scanned contracts, receipts, or filings, such as Ouria intensive receipts or IFPAS vendor quotes.

In practice: A vendor confidently publishing its own benchmark, worth reading as a vendor benchmark rather than a neutral one.

For: Both runtimes. A Python library and hosted API with no Claude Code specific hook, usable from Axion or Hermes alike.

Security3

Quality4

Auditability3

Useful to you2

Useful to community4

Buildable now3

Hermes2

Verdict: watch. A plausible future fit, but no current Axion workflow is blocked on document extraction today, so this is worth tracking rather than building now.

Build #11 LlamaIndex Extract v2.5: use the ai-implementation-build-intake skill to build this safely. Source: https://www.llamaindex.ai/blog/introducing-extract-v2-5. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · ExtractBench scores 14 systems across 370 documents. Extract v2.5 was announced through LlamaIndex's official blog.

#12

newwatchBoth runtimes21 / 35

Mastra, mastra core 1.75.0 repo

Mastra's core package reached 1.75.0 on October 7, 2026, continuing a release cadence that shipped 6 versions, 1.70 through 1.75, in the first week of October alone.

What it does for you: A TypeScript first agent framework that would fit closer to any future Hermes web tooling than the Python first frameworks this registry mostly tracks, though nothing in YY's current workload is blocked on it today.

In practice: Fast moving in a way that reads more like a startup shipping daily than a framework settling into stability.

For: Both runtimes. A TypeScript agent framework with no Claude Code specific binding, usable from any Node based Axion or Hermes component.

Security3

Quality4

Auditability3

Useful to you2

Useful to community4

Buildable now2

Hermes3

Verdict: watch. No concrete gap this fills over Claude Code's native subagent system today, but the release cadence and TypeScript focus are worth tracking for future web facing agent work.

Build #12 Mastra, mastra core 1.75.0: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/mastra-ai/mastra/releases/tag/%40mastra%2Fcore%401.75.0. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · mastra core 1.75.0 released October 7, 2026. 6 releases shipped between October 1 and October 7, 2026.

#13

newwatchBoth runtimes20 / 35

CrewAI 1.15.24 repo

CrewAI shipped version 1.15.24 on October 7, 2026, with a GitHub verified signed commit, the latest patch in its 1.15.x line for its role based multi agent orchestration framework.

What it does for you: A well maintained orchestration framework, but one that overlaps with Claude Code's native subagent and workflow system Axion already runs on, so it is not a clear gap filler today.

In practice: Mature and steady, the kind of project that keeps shipping patches without drama.

For: Both runtimes. A Python library with no Claude Code specific binding, usable from Axion or Hermes alike.

Security3

Quality4

Auditability3

Useful to you1

Useful to community4

Buildable now3

Hermes2

Verdict: watch. Same basis as this registry's prior calls on CrewAI: a well maintained orchestration framework that duplicates what Claude Code's native subagent system already covers for how this workspace operates.

Build #13 CrewAI 1.15.24: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/crewAIInc/crewAI/releases/tag/1.15.24. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Version 1.15.24, released October 7, 2026 with a GitHub verified signed commit.

#14

newwatchClaude (Axion)20 / 35

Claude Marketplace, 2,000 plus plugins and connectors plugin

Anthropic's own blog confirms the Claude Marketplace has crossed 2,000 plugins and connectors, bringing products, agents, and service partners into one discovery surface built on MCP and Agent Skills.

What it does for you: A single place to re check before building anything new, since the enabledPlugins read only check this skill runs every day depends on this same marketplace. None of YY's 14 currently enabled plugins changed because of this milestone.

In practice: Reads as Anthropic positioning Claude as a platform with its own app store, not just a product.

For: Claude (Axion). The marketplace itself is a Claude Code and claude.ai specific distribution surface, with no Hermes equivalent.

Security2

Quality4

Auditability2

Useful to you3

Useful to community5

Buildable now3

Hermes1

Verdict: watch. A real milestone worth tracking, but a marketplace of 2,000 plus plugins and connectors means most entries get essentially no individual vetting, so it is a place to check, not a thing to adopt wholesale.

Install: browse claude.com/platform/marketplace and review each listing's publisher and permissions individually before installing anything. Review before enabling. Copy

Source · 2,000 plus plugins and connectors, reported October 4, 2026 and confirmed on Anthropic's own blog.

#15

newwatchBoth runtimes19 / 35

Google ADK (Python) v2.11.0 repo

Google's Agent Development Kit for Python reached v2.11.0, the latest in its post 1.0 release line for building and deploying multi agent systems on Google's stack.

What it does for you: Same overlap reasoning as CrewAI applies, plus it ties YY toward Google's own agent stack rather than the Claude native one Axion already runs on.

In practice: Solidly maintained and clearly Google's answer to CrewAI and LangGraph, not a gap filler for this workspace.

For: Both runtimes. A Python library with no Claude Code specific binding, though its deployment path leans toward Google Cloud and Vertex.

Security3

Quality4

Auditability3

Useful to you1

Useful to community4

Buildable now2

Hermes2

Verdict: watch. Same framework overlap reasoning as CrewAI; ADK's deployment path leans toward Google Cloud and Vertex, a second platform dependency YY does not currently need.

Build #15 Google ADK (Python) v2.11.0: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/google/adk-python/releases/tag/v2.11.0. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Version v2.11.0 confirmed as the latest tagged release as of October 7, 2026.

#16

newwatchStandalone tool19 / 35

Cline v4.1.23 repo

Cline, a model agnostic IDE coding agent with over 30 providers and parallel agent execution with Kanban task boards, shipped application version v4.1.23, a separate release line from the SDK release, v0.0.88, this registry already logged on October 4, 2026.

What it does for you: A capable competitor to Claude Code itself, but adopting a second IDE coding agent alongside Claude Code would fragment YY's workflow across two tools with separate configs and separate learned habits.

In practice: A serious, actively developed competitor, not a toy, but still a competitor rather than a gap filler.

For: Standalone tool. A standalone VS Code extension and coding agent product that competes with Claude Code directly, not a library Axion or Hermes would embed.

Security3

Quality4

Auditability3

Useful to you1

Useful to community4

Buildable now2

Hermes2

Verdict: watch. Same framework overlap reasoning as CrewAI and ADK; a second IDE coding agent alongside Claude Code would fragment YY's workflow rather than fill a gap.

Build #16 Cline v4.1.23: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/cline/cline/releases/tag/v4.1.23. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Version v4.1.23 is the current latest tagged application release, logged after the SDK v0.0.88 update this registry already captured on October 4, 2026.

#17

newwatchStandalone tool19 / 35

Goose, Linux Foundation AAIF governance transfer repo

Goose's repository moved to the aaif-goose GitHub organization, reflecting its transfer to the Linux Foundation's Agentic AI Initiative foundation governance, with recent pull requests covering MCP version preference, ACP client packaging, and removal of its managed model registry in favor of direct provider connections.

What it does for you: Foundation governance plus a move toward direct provider connections makes Goose a lower lock in open source coding agent option, worth tracking even though YY's primary tool remains Claude Code.

In practice: A governance change more than a feature change, the kind of move that matters for long term trust, not day to day usefulness.

For: Standalone tool. A standalone CLI coding agent product that competes with Claude Code directly, not a library Axion or Hermes would embed.

Security3

Quality4

Auditability3

Useful to you1

Useful to community4

Buildable now2

Hermes2

Verdict: watch. Same framework overlap reasoning as Cline and CrewAI; adopting it would duplicate Claude Code's role rather than fill a gap, though the governance move itself is worth noting.

Build #17 Goose, Linux Foundation AAIF governance transfer: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/aaif-goose/goose. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Repository now under aaif-goose/goose. Pull requests 11827, 11846, and 11789 landed in the most recent release.

#18

newwatchStandalone tool19 / 35

Codex CLI 0.161.0 repo

OpenAI's Codex CLI reached version 0.161.0 as its current latest tagged release, with the rust-v0.161.0-alpha.13.1 pre release landing October 6, 2026, continuing a roughly weekly release cadence in its Rust based CLI line.

What it does for you: Same framework and tool overlap reasoning as Cline and Goose; a third free coding agent CLI alongside Claude Code does not fill a gap YY has today.

In practice: A credible, fast moving competitor from a major lab, worth watching for parity pressure on Claude Code rather than for direct use.

For: Standalone tool. A standalone Rust CLI coding agent product from OpenAI that competes with Claude Code directly, not a library Axion or Hermes would embed.

Security3

Quality4

Auditability3

Useful to you1

Useful to community4

Buildable now2

Hermes2

Verdict: watch. Same framework and tool overlap reasoning as Cline and Goose; a patch level release carries no new capability worth evaluating against Claude Code's native tooling.

Build #18 Codex CLI 0.161.0: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/openai/codex/releases/latest. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Version 0.161.0 is the current latest tagged release, with the rust-v0.161.0-alpha.13.1 pre release dated October 6, 2026.

#19

newwatchBoth runtimes15 / 35

Pensyve, Cloud shutdown and maintenance mode repo

Pensyve, an external Claude Code marketplace entry for universal agent memory, closed its hosted Cloud offering on October 1, 2026. The open source engine, SDKs, and documentation continue under the Apache 2.0 license, now in maintenance mode with security fixes and dependency updates only, no new features planned.

What it does for you: A caution more than an opportunity. If YY or any future Axion build had been eyeing Pensyve Cloud for memory hosting, that path is now closed, self hosting is the only option, and the project itself says not to expect new features.

In practice: A quiet shutdown notice rather than a dramatic one, the kind of status update that is easy to miss if nobody reads the README.

For: Both runtimes. A self hostable, open source memory runtime with no Claude Code specific binding, though its hosted Cloud offering just closed.

Security3

Quality2

Auditability3

Useful to you1

Useful to community2

Buildable now2

Hermes2

Verdict: watch. The project is explicitly in maintenance mode with its hosted Cloud product shut down, so this is worth noting as a risk signal rather than evaluating as something to adopt.

Build #19 Pensyve, Cloud shutdown and maintenance mode: use the ai-implementation-build-intake skill to build this safely. Source: https://github.com/major7apps/pensyve. Save canonical skill/agent under AXION\Skills and AXION\Agents. Copy

Source · Pensyve Cloud closed October 1, 2026. The open source engine remains under the Apache 2.0 license.

#20

newwatchClaude (Axion)14 / 35

maddog plugin

An advisor executor agent universe for Claude Code built around a judgment tiered executor family, a product engineering pipeline, and gated authoring and session mining skills, submitted to awesome-claude-code on October 7, 2026.

What it does for you: A structured multi tier agent delegation pattern, advisor then executor, that overlaps with but formalizes further than Claude Code's native subagent system. Worth a careful look for how it compares to Axion's own agent hierarchy rules, not an immediate adoption.

In practice: An ambitious, large scope plugin that reads as one person's whole agent philosophy packaged at once, which is exactly why it needs a careful read before use.

For: Claude (Axion). Ships as a Claude Code plugin with a .claude-plugin folder and Claude specific agent and skill directories.

Security2

Quality2

Auditability2

Useful to you3

Useful to community2

Buildable now2

Hermes1

Verdict: watch. A full agent universe plugin submitted the same day as this scout run, with session mining and gated authoring components that need a careful read before pointing at any real Axion workspace, given this skill's own two level hierarchy rule.

Install: run claude plugin marketplace add Harish-here/maddog, then install only the specific executor tier needed rather than the full agent universe. Review before enabling. Copy

Source · Submitted to awesome-claude-code on October 7, 2026.