Vesara
Vesara Daily Sunday, July 19, 2026
 
The practical agent story is shifting from one model to a routing layer
Today at a glance
Every published a useful look at the ops work behind routing tasks across models, while the strongest operator signals today are observability, local research, and tighter control over agent context. Funding is thin: the clearest fresh European round is Greenjets at €35m.
 
01  Agent abilities Skills · MCPs
 
Skills
01 Prototype
A hot Skills.sh package for rapidly turning a rough product idea into a working prototype.
Why it matters: Use it to shorten the distance between a Vesara workflow idea and a testable artifact.
Skills.sh Product build High
02 Lark Task
A hot integration skill for creating and managing tasks in Lark.
Why it matters: Useful if client delivery moves through Lark and agents need to leave durable work behind.
Skills.sh Workflow Medium
03 AI Music
A trending skill for AI-assisted music work, from generation through basic production tasks.
Why it matters: Worth watching as a pattern for packaging specialist workflows, not as a priority install.
Skills.sh Creative Low
04 Workflow automation
A workflow orchestration skill that covers creating, executing, and templating multi-step agent processes.
Why it matters: Relevant for turning a proven lead research process into a repeatable, inspectable runbook.
SkillsMP Orchestration High
05 Persistent memory
A memory skill focused on preserving a task ledger and decision history between sessions.
Why it matters: Good fit for long-running client agents where dropped context becomes operational debt.
SkillsMP Memory High
 
MCPs
01 Keenable Web Search
A verified remote web-search MCP with ranked results, date and site filters, plus clean markdown fetching.
Why it matters: A compact research primitive for agents that need current web evidence before acting.
Smithery Research High
02 FavCRM
An agentic CRM MCP for service businesses with typed tools across customers, bookings, loyalty, invoicing, and WhatsApp or SMS.
Why it matters: Interesting for client agents that need to update operating systems, not merely draft a follow-up.
Smithery CRM High
03 OneSignal
A customer-engagement MCP for push, email, SMS, and in-app campaigns, including audience management and reporting.
Why it matters: Useful when an agent should launch and measure a lifecycle action from the same workflow.
Smithery Engagement Medium
04 Brave Search
A verified MCP for Brave’s independent web index across web, news, images, and video.
Why it matters: A sensible alternative search source when research agents need coverage beyond one index.
Smithery Research Medium
05 DevMatch
A hiring-oriented MCP that searches engineers by verified open-source work rather than profile keywords.
Why it matters: Potentially useful when Vesara needs a proof-of-work layer for technical candidate or partner sourcing.
Smithery Talent Medium
 
02  Trending repos GitHub · last 24h
 
01 SigNoz/signoz  MIT — An OpenTelemetry-native observability platform with logs, metrics, traces, and an AI-agent focus. (+432 today) 30,841 ★
02 KnockOutEZ/wigolo  No licenseearly — A local-first search, fetch, crawl, and research layer designed for AI coding agents over MCP. (+203 today) 1,383 ★
03 apache/ossie  Apache-2.0early — Apache’s effort to standardize semantic metadata across analytics, AI, and BI tools. (+47 today) 1,324 ★
04 sourcebot-dev/sourcebot  Apache-2.0 — A self-hosted codebase understanding tool for humans and agents. (+16 today) 3,618 ★
05 musistudio/claude-code-router  MIT — A local control plane for routing agent work across models and tools. (+31 today) 35,899 ★
06 upstash/context7  MIT — A platform for feeding current library documentation into LLMs and coding tools. (+69 today) 59,392 ★
Trending ≠ vetted. Star counts measure attention, not safety — review the code and pin versions before running anything marked early.
 
03  AI & tech news Key reads · max 5
 
01 The ops team that routes work across models  Every
Every’s new field note focuses on the people and rules behind model routing.
02 Claude Code uses Bun written in Rust now  Simon Willison
Claude Code versions after 2.1.181 reportedly use Bun’s Rust port.
03 Prompt injection attacks are thwarting AI hacking agents  Wired
Wired describes context-bombing attacks that can make autonomous security agents stop or lose the thread.
04 How we built Gift Links  Every
Every’s editorial team explains how it shipped a website feature with AI assistance.
 
04  Reddit watch Top 5 · practitioner signal
 
01 Agentglass: a view into what Claude is doing  R/CLAUDEAI
A Claude Code user built a dashboard because tmux panes did not answer the operational questions: which session is stuck, what changed, and where time is going.
02 Discussion: creator bias in LLM answers  R/CLAUDEAI
A community thread raises the uncomfortable possibility that models favor their own makers.
03 Gmail MCP one-line install  R/MCP
A developer shared a lean Gmail MCP setup intended to reduce installation friction.
04 DeepSeek efficiency discussion  R/LOCALLLAMA
Local-model users are dissecting DeepSeek’s price-performance profile against newer competitors.
05 Kimi K3 on SpreadsheetBench 2  R/LOCALLLAMA
A thread points to Kimi K3’s SpreadsheetBench result.
 
05  Funding Pre-seed · Series · Growth
 
01 Greenjets · €35m  Series A
The UK aerospace company raised a €35m Series A led by Blossom Capital, with participation from the NATO Innovation Fund and UK institutions, for propulsion and aircraft-platform work.
02 Lakestar Resilience I · €262.2m  Fund close
Lakestar closed a €262.2m fund dedicated to European defence and dual-use companies.
Fresh operator-relevant financings were sparse today; these are the two credible European capital signals that cleared the bar.
 
06  Research watch HF Papers · weekly top · max 5
 
01 Ring-Zero: Scaling Zero RL to a Trillion Parameters  HF Papers · 90 upvotes upvotes · Jul 14
The paper studies reinforcement learning with verifiable rewards at trillion-parameter scale.
02 VideoChat3: Fully Open Video MLLM  HF Papers · 123 upvotes upvotes · Jul 16
VideoChat3 is an open video model aimed at long video, streaming interaction, and general video understanding.
03 Boogu-Image-0.1  HF Papers · 125 upvotes upvotes · Jul 14
Boogu-Image-0.1 is an open multimodal family for image understanding, generation, and instruction editing.
 
Vesara mark Vesara Post-AI. Human-native.
Vesara Daily · Curated by Vesara operators · © 2026 Vesara, Inc.