A native Mac notes app that keeps notes on the screen edge and lets an agent search, write, and communicate with notes from any application.
A notepad connected to AI assistants that saves conversation outputs as searchable text, Markdown, or HTML clips with sharing and version history.
A Slack bot that builds an organization chart by asking every person who they report to and updates the chart as the organization changes.
AIRUNCODE is a local-first agent runtime for running coding agents on a user's own computer. It supports voice-driven code-agent tasks and access to AI providers at their origin pricing, while its site describes execution as local.
AI Toolbox is a Chrome extension, formerly called ChatGPT Toolbox, that adds folders, subfolders, full-text search, prompt management, conversation export, and other organizational features to ChatGPT, Gemini, Claude, and Grok. Its cross-platform search indexes conversations locally, groups results by AI service, and links directly to the matching chat. It also supports saved prompts, a curated prompt library, sequential prompt chains of up to 10 steps, bulk conversation actions, smart tags, bookmarks, audio downloads, and exports in text, Markdown, JSON, PDF, or ZIP formats. The extension runs on Chrome, Edge, Brave, and other Chromium browsers; the site describes conversation content as remaining on the user's device, while folders, prompts, and tags can sync in encrypted form. A free plan is available, with paid plans and team features.
BrickForgerAI is a web tool that turns text prompts into custom, buildable brick models. It generates a previewable 3D design from a library of real, purchasable brick parts, showing the part count and dominant colors before purchase. Users can download an LDraw .ldr file, a full parts list, and step-by-step PDF building instructions. The service offers small, medium, and large build sizes and currently performs best with organic shapes such as animals, plants, and sculptures. Its stated library contains 55 part types and is being expanded with techniques such as sideways building (SNOT); the site says generated models commonly achieve full connectivity but may receive stability warnings in BrickLink Studio. BrickForgerAI is not affiliated with LEGO, BrickLink, or Studio.
BrickForgerAI is an AI service that generates custom buildable brick models from text prompts. It previews each model in 3D with its part count and dominant colors, then provides a downloadable .ldr file, full parts list, and step-by-step PDF building instructions for purchased generations. The service generates models from a library of 55 real, purchasable brick parts, including slope angles and curved pieces. Its site says it is currently strongest at organic shapes such as animals, plants, and sculptures, and is adding techniques including sideways building (SNOT). Build sizes are offered at approximately 15, 22, or 30 studs; the listed plan costs £1.50 per month for three generations with the .ldr file and instructions included. Part geometry comes from the LDraw parts library under CCAL 2.0. BrickForgerAI states that it is not affiliated with LEGO, BrickLink, or Studio.
Clipnote is a notepad for saving and organizing content generated in AI conversations. It stores plain text, Markdown, and HTML clips, preserving HTML layout and interactivity, and lets users search, pin, archive, version, restore, and delete them. Clips can be private or shared through public URLs, and collections bundle multiple clips into one shareable page. Through MCP connections with Claude and ChatGPT, an assistant can save a new clip, retrieve a clip's content to inform a response, or update an existing clip. Updates automatically preserve the previous version. Clips can also be created or updated manually by pasting text or uploading a file.
Compliance by TwelveLabs is an AI video-compliance application that applies multimodal retrieval and contextual reasoning to video libraries against compliance rules that users author, version, and control. It identifies and explains potentially problematic moments, provides timestamped evidence and review workflows, and exports signed reports through a managed service and REST API.
An AI video-compliance product from TwelveLabs that reviews video libraries against configurable regional or custom compliance rule packs. It produces evidence and signed reports for the results.
dif.sh is a developer tool for storing feature flags, A/B tests, holdouts, and staged rollouts as Markdown files in a Git repository. Each experiment uses frontmatter for its status, audience, variants, metrics, guardrails, and exclusion group, while Git provides versioning and pull requests provide review and audit history. Its CLI initializes the repository structure, drafts experiments from surface-specific learning logs, validates configurations, previews assignments, builds a typed client, and archives concluded tests. The build resolves exclusion groups and fails on conflicting live experiments, generates a client that evaluates variants in the application without a production service call, and refreshes dif/context.json for coding agents. Audience rules declare runtime attributes such as country, plan, or returning-visitor status without committing customer lists. The optional Dif Cloud integration can calculate lift through dif.track(); alternatively, custom event handlers can forward results to other analytics systems. The project provides an npm CLI package and an SDK, and its site describes a no-signup workflow.
Experiential Labs is an open-source AI gateway that provides an OpenAI-compatible endpoint for hosted model providers, customer-owned provider keys, private GPUs, and self-hosted models. It routes requests through a single API and key while handling provider access, model selection, failover, and streaming responses. Its optional intelligence layer monitors traffic to identify model switches, improve cache hit rates, and route requests to customer-owned fine-tuned models. The service also provides organization-wide usage attribution, request logs, spend reporting, access controls, model allowlists, and per-key spending caps scoped by team, person, agent, or tool. It can be used as a hosted gateway or self-hosted, with hosted inference and Pro offered alongside the open-source gateway.
fal is a generative media platform for developers that provides APIs and SDKs for running image, video, audio, and 3D models. It offers a gallery of production-ready models, including open models and user-provided LoRAs, accessible through a unified API. The platform runs inference on serverless GPUs and can scale deployments from zero to large numbers of GPUs. It also provides compute for training and fine-tuning, dedicated clusters for custom workloads, private model endpoints, and observability tools for monitoring deployments and usage.
GitWarren is a desktop application for reviewing local Git changes made by coding agents before they are committed. It reads the worktree directly and combines staged, unstaged, and untracked files into one diff for local review, with comments and threaded discussions that can be edited or resolved. An MCP server over stdio lets Claude Code, Codex, or another MCP client open reviews, inspect discussions, reply in threads, comment on lines, and resolve comments; agent messages are attributed and separate MCP sessions distinguish concurrent agents. Reviews and comments are stored in a local SQLite file, while repository state is read from Git when displayed. The application requires no account, runs on macOS, Windows, and Linux, and is free and open source under GPL-3.0.
HyperProbe is an AI-native production debugger for investigating live application state without code changes, redeployment, or service restarts. It reads logs and distributed traces, uses a coding agent to locate a suspected file and line, and places a temporary read-only virtual breakpoint there. When the breakpoint fires on live traffic, it captures the variable state asynchronously without pausing requests, then uses the evidence to confirm a root cause. Probes are non-blocking, cannot write memory or execute code, disappear after capture, and are recorded in an immutable audit trail. The service is designed for silent failures, exceptions whose causes occur earlier in the call chain, incorrect behavior without thrown errors, race conditions, third-party contract changes, and business-metric failures. It supports JavaScript, TypeScript, Java, Kotlin, Python, and Ruby, and can run in a managed cloud, self-hosted deployment, or private VPC. The page states that it integrates with PagerDuty, Datadog, Slack, Cursor, Claude Code, Codex, and Opencode, with approval gates and agent-side PII redaction.
Lightfield is an AI-native customer relationship management (CRM) platform for early-stage teams, designed to capture, analyze, and surface information about customers and interactions using AI-driven features.
Ponytail is a rule set and plugin for AI coding agents that directs them to implement the smallest working solution. Its decision ladder skips speculative requirements, reuses code already in the repository, prefers the standard library and native platform features, avoids adding dependencies, and then chooses the shortest implementation that works. The project says that validation, error handling, security, and accessibility are not simplified away. It provides chat commands for changing enforcement intensity, reviewing the current diff for over-engineering, auditing an entire repository for bloat, recording deferred shortcuts in a debt ledger, and viewing benchmark results. It can be installed across multiple coding-agent environments, including Claude Code, Codex, Copilot CLI, Gemini CLI, and Pi; the page also lists support for other agents.
Reflexio is a learning platform for AI agents that turns user corrections, failed paths, and successful outcomes into reusable behavioral changes without retraining the underlying model. Its SDK and integration loop publish interaction outcomes, extract actionable feedback, store learned behaviors, and retrieve only relevant learnings during later inference; integrations are available through Python, REST, a CLI, and a portable coding-agent skill. Reflexio evaluates learnings against control responses and user-defined success criteria, tracking whether they solved the user's problem, required correction, or escalated to a human. It makes learnings auditable and revocable: they can be reviewed, rewritten, approved, rejected, or deleted, with rejected learnings removed from retrieval. Background processes consolidate duplicate signals and resolve conflicts or outdated lessons, while evidence from later sessions is used to revise learnings. The service supports managed, bring-your-own-key, customer-owned database, bring-your-own-cloud, and fully self-hosted deployments. The page identifies an official repository at github.com/ReflexioAI/reflexio and states that the same API can be used across deployment modes.
Routines by Databox is a scheduling feature for recurring analytics and reporting. Users save an analysis as a reusable Skill, schedule that Skill to run daily, weekly, or monthly, and receive the results by email, Slack, or in Databox. Databox describes Routines as running Skills automatically and delivering the results, alongside AI agents that can delegate broader analysis-to-action workflows under user oversight.
Scriptly is an iOS teleprompter app that uses real-time voice recognition to scroll a script at the speaker’s pace. It provides script writing and organization, video-recording controls, professional teleprompter tools, local storage, and an Apple Watch companion for recording and text adjustments. The public beta is distributed through TestFlight.
Snitch is a Slack bot that builds and maintains an organization chart by asking each workspace member who they report to. An administrator connects it through Slack OAuth, reviews the people to include and the exact message, and launches the question round; members answer in a DM, and can later update their reporting line. The bot answers questions about reporting chains, team sizes, and ownership in Slack, while flagging problems such as missing or departed managers, reporting loops, and multiple top-level executives. It uses the Slack member directory and messages sent to it, not channels, files, or other conversations. The service is offered as a subscription with a 14-day trial and workspace-size-based plans.
Speakeasy is an AI control plane for discovering, securing, and governing enterprise AI agents, MCP servers, Skills, and AI applications. It maintains a catalog of approved AI tools, gives each agent an identity through the organization's existing identity provider, and scopes access by team and role. Every prompt, response, and tool call passes through the control plane for real-time policy checks before reaching internal APIs, MCP servers, or SaaS tools. Policies can distinguish read and write access and individual tools; violations are blocked, while allow-or-deny decisions are recorded in an audit trail. The platform is designed to detect and quarantine unapproved MCP connections and to block threats such as prompt injection, PII exposure, and leaked credentials in flight. Speakeasy deploys through an organization's existing MDM, including Jamf and Intune, and integrates with SAML/OIDC identity providers. The page states that it is SOC 2 Type II audited, ISO 27001 certified, GDPR compliant, and HIPAA ready.
Tadata is an AI employee for Slack that handles research and repetitive sales and go-to-market operations work. It can prepare call briefings, research prospects and companies, draft outreach and follow-up emails, fill CRM fields, monitor relevant LinkedIn activity, and suggest recurring automations. It connects to tools such as HubSpot, Attio, Notion, Linear, GitHub, Google Sheets, Gmail, and Google Calendar, as well as external web sources and MCP or API integrations. Tadata runs pre-built agents or builds an agent from a described process. It observes repeated work, learns team preferences, and requests approval before automating tasks or sending messages; connected-tool access is scoped to the user's authorization. The service is presented as model-agnostic, with portable automations, preferences, exceptions, and agent behaviors that can be exported and versioned.
Tucky is a native macOS notes app that docks notes as a thin stripe on the edge of the screen. Notes remain local and encrypted on disk, with Markdown, checkbox tasks, exports, global shortcuts, and local-only files. Its integrated AI agent can search, write, rewrite, and talk to notes from the edge without opening another window. The app supports voice dictation and connectors for Gmail, Google Calendar, Google Docs, Google Sheets, GitHub, and Notion; a shortcut can ask about the current note while limiting whether its body is read. The free tier supports five notes without AI, voice, or connectors, while the Plus plan adds unlimited notes and those features. Tucky requires macOS 15 or later and supports Apple silicon and Intel.
Tucky is a native macOS notes app that keeps notes in a thin stripe at the edge of the screen. Notes are stored locally and encrypted on disk, with support for live Markdown, checkbox tasks, global shortcuts, and clean exports. Its optional AI features let users search, write, rewrite, and ask questions about notes from the edge or through the global ⌥Space shortcut. The app can connect to Gmail, Google Calendar, Google Docs, Google Sheets, GitHub, and Notion; voice dictation is supported in more than 60 languages. The page says note titles accompany questions by default, while note bodies are read only when allowed. The free plan supports five notes without AI, voice, or connectors. The Plus plan is listed at $4 per month and adds unlimited notes, AI models, voice, and connectors. Tucky runs on macOS 15 or later on Apple silicon or Intel.
Searchable transcript of Top AI Agent Projects : Tucky, Airuncode, Tadata, HyperProbe & GPT-6 Astra — ManuAGI - AutoGPT Tutorials (22:28). Search for a phrase, then click its timestamp to jump straight to that moment in the video.
Captions sourced from the original video on YouTube, published by ManuAGI - AutoGPT Tutorials. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.
00:00 Everyday developers build new AI tools and keeping up with all of them takes time. This is our weekly project update video where we cover the top trending AI agent projects this week so you can find what is worth using in one place. You will discover trending AI tools and projects across notes, coding, and automation without wasting time. Let's get started.
00:21 >> Before we jump into today's project updates, here's a quick announcement for everyone. We've launched a brand new YouTube channel called AI Agent Studio dedicated entirely to AI Agent projects, tutorials, and tools. So, if you're interested in staying up to date with the latest AI agent open source projects, learning how to build your own agents, or exploring cuttingedge agent frameworks, make sure to check it out.
00:47 Subscribe now to get weekly videos, in-depth guides, and realtime project breakdowns. The link is right there in the description. Don't miss it. All right, let's get into today's video. >> Project number one, Tucky Edge Macotes with an agent. Tucky is a native Macotes app that lives as a thin stripe on the edge of your screen with an agent built in.
01:08 It solves the way notes and AI usually sit in another window you have to go find. Notes dock to the edge. Titles turn sideways so you can read them. And you reach over to fan them out and open one. You hold a shortcut and ask from any app. So the agent searches, writes, and talks to your notes. And the best models are already in the app with nothing to configure.
01:28 Titles go with a question, but the body is read only when you allow it. Everything stays encrypted on your Mac in a local only file with voice dictation in over 60 languages and connectors for Gmail, Calendar, Docs, Sheets, GitHub, and Notion. It runs on Apple Silicon and Intel. It is built for Mac users who take notes all day. Download it and keep one note free.
01:51 Project number two, lost a clip note. Save AI conversations as sharable clips. Clipote is a notepad that connects to your AI and keeps the work your conversations produce. It solves the way chat GPT and clawed answers vanish the moment you close the tab, leaving nothing to return to. You tell your AI to save this and the reply is kept on the spot as a clip in plain text markdown or HTML shown exactly as it looked and pasting or dropping in a file works too.
02:20 It connects over MCP with one button on Claude AI and developer mode in chat GPT. No keys needed, so the AI browses your clips, pulls one into a reply, saves a new one, or updates an existing one by name. HTML clips keep their layout and interactivity. So, a bill splitter or a small game stays live behind a URL. You keep clips in a searchable library, bundle them into collections, and share any as a private or public link with version history and a 30-day trash.
02:50 It is built for anyone who works in AI chats. Get started free and save your first clip. Project number three, Snitch Slack org chart that builds itself. Snitch is a slackbot that builds your org chart by asking each person one question and keeps it true as people move. It solves the stale chart problem where past 20 people, the real reporting lines live in an old slide, a spreadsheet, and whoever is online to ask.
03:16 You add it through Slack in one click. Exclude anyone you want and approve the exact message. Then it DMs everyone, who do you report to? And each person answers in two taps. The chart draws itself from answers people give for themselves. Nothing inferred from a profile and it comes back to ask again when someone joins. A line changes or a manager leaves.
03:36 You tag it in any channel to ask a reporting chain, a team size or who owns a thing in plain language and it flags loops, empty chairs and a stray second CEO instead of going quietly wrong. It reads only the DMs sent to it and the chart is visible only to your verified workspace. It is built for growing teams. Add it to your Slack. Free project number four.
03:57 Aaron code. Local first agent runtime with voice. Aaron code is a local first agent runtime that lives on your own machine. It solves the privacy and cost trade-offs of coding agents that run in the cloud and markup model usage. It runs code agents entirely on your computer, so execution stays local and your work does not leave the machine. You drive those agents by voice, speaking a task instead of typing it.
04:24 and reach every major AI provider through it at origin pricing with no markup on what you spend. It is built for developers who want private hands-free coding on their own hardware. Try it and run your agents locally. Project number five, routines by datab box. Schedule recurring analytics to run itself. Routines is a scheduling layer in databox as a gentic analytics platform that runs your recurring analysis on autopilot.
04:48 It solves the drag of pulling the same reports by hand, where the insights a team relies on depend on someone remembering to build them. First, you capture how you work as a skill, saving the steps, context, and standards behind an analysis so anyone can run it exactly the way you would or grab one from the skills marketplace. A routine then puts that skill on a schedule daily, weekly, monthly, or a cadence you set, and delivers the finished report to email, Slack, or inapp with the subject line flagging whether
05:18 anything needs you. Run history keeps every past run and its report, and you ask follow-up questions in chat. It works on over 130 integrations, plus context pulled through MCP connections with agents that carry a whole workflow coming next. It is built for agencies and growing teams. Try it free and automate your first report. Project number six, Skipley, voice following teleprompter for faster filming.
05:42 Scrippley is an iOS teleprompter app that brings a creator's whole recording workflow into one place. It solves the juggling of notes, timing, and endless retakes, and the awkward chase of a teleprompter set to a fixed speed. As you speak, real-time voice recognition follows your words and scrolls at your own pace, so you never race the text or fall behind.
06:02 You write and organize your lines ahead of time and keep them ready. Then film with the prompter running and an Apple Watch companion starts and stops recording and adjusts the text without reaching for your phone. Your content stays in local storage on your device for privacy. It suits vertical short form for social platforms. It is built for creators who record often, download the public beta, and record faster.
06:26 Project number seven, AI Toolbox 3.0 folders and search for AI chats. AI Toolbox is a Chrome extension that adds folders, search, a prompt library, and export inside Chat GPT, Gemini, Claude, and Grock. It solves the flat, unarchable sidebar these tools ship with, where hundreds of chats pile up with no structure and no way to find what was actually said.
06:50 It loads into the chat you already use and puts its tools in the sidebar. So you sort conversations into nested folders, pin the ones you return to, and run one full text search across all four platforms at once, grouped by AI with deep links. Smart tags categorize chats on your device. You save prompts and call them with a shortcut or chain several to run in order, and you export single or bulk conversations as text, markdown, JSON, or PDF.
07:18 It works on Chromium browsers with no account. Your conversation content stays in your browser and only your folders, prompts, and tags sync. Encrypted. It is built for people who live in AI chats. Add it to Chrome. Free project number eight. DIA TA AI employee in Slack that acts. Data is an AI employee that lives in Slack and takes research and repetitive work off your plate.
07:42 It solves the busy work that eats a go-to market team's day where prep, follow-ups, and CRM updates pile up between calls. It runs in your tools around the clock and pings you only when there is something to act on. Briefing you each morning on your calls and what changed, drafting follow-up emails, filling CRM fields, and writing the next step after a call with nothing sent until you approve.
08:04 It learns how you like things done, notices work that repeats, and offers to automate it, asking first and seeing only the tools you connect. It reaches across HubSpot, Adio, Notion, Linear, GitHub, Gmail, and more through any MCP or API and pulls external context from company sites, LinkedIn, job boards, and filings. It stays model agnostic, so your recipes and preferences are yours to export and take between providers with scoped access and isolated credentials.
08:33 It is built for sales and GTM ops teams. Get started free with credits included. Project number nine, fail AI generative media platform for developers. Fail.AI is a generative media platform that lets developers run image, video, audio, and 3D models through one API. It solves the setup and infrastructure burden of using these models where each would otherwise mean its own GPUs, scaling, and mops.
09:02 You call a unified API or SDK to reach over a thousand productionready models. No fine-tuning or configuration needed or bring your own weights and loras. Its inference engine is tuned for fast diffusion, so you scale from a prototype to hundreds of millions of daily calls without cold starts or autoscaler work. And serverless GPUs spin from zero to thousands on demand.
09:23 For heavier work, dedicated clusters on current NVIDIA hardware handle training and custom models with one-click private endpoints and observability built in. You pay per output on serverless or hourly on compute with no lockin. It is built for developers and companies shipping AI media features. Get started free and make your first call. Project number 10, Speak Easy.
09:46 AI control plane governing every agent. Speak Easy is an AI control plane for discovering, securing, and governing the AI agents, MCP servers, and skills across an enterprise. It solves the shadow AI problem where security and identity tools built for people signing into apps break down once agents reach production on borrowed credentials and shared accounts.
10:07 It keeps one catalog of approved agents, servers, and skills so you can see how AI is used and it extends the identity provider you already run so every agent acts under its own identity with access scoped by team and role. It deploys through your existing device management and sits on the path between agents and your systems, checking every prompt, response, and tool call against policy in real time, letting allowed actions through and stopping prompt injection, PII exposure, and leaked credentials in flight with each
10:38 decision logged for audit. It works with Claude, Chat, GPT, Cursor, and any MCP agent, and runs in your cloud or on premises. It is built for enterprise security and platform teams. Explore the platform or talk to the team. Project number 11, diff feature flags and AB tests in Git. Diffes is a developer tool that keeps feature flags and AB tests as markdown files inside your repo.
11:03 It solves the split between code and git and experiments in a separate dashboard where flags drift conflict and lose the reasoning behind them. Each flag or test is one diff MD file with front matter. So a flag, an AB test, a hold out, and a staged rollout share the same shape, versioned in Git and reviewed in a pull request. You declare which audience attributes to target, and real values arrive at runtime, so no customer list is committed.
11:31 A build step compiles a typed client your app imports directly, refuses to compile when two live tests would collide on one user, and regenerates a context file your coding agent reads on session start, like a memory of active tests and past learnings. A CLI drafts, validates, previews, and concludes each test, archiving it, and logging what it taught so the next one starts informed.
11:53 An optional cloud layer reads the repo for a teamwide view and proposes verdicts as pull requests. It installs in one command with no signup. It is built for developers and their agents. Install it and put your next flag in Git. Project number 12, Reflexio, learning platform that improves your agents. Reflexio is a learning platform that turns static AI agents into self-improving systems.
12:17 It solves the way ordinary agents repeat the same mistakes. Since a memory layer stores what a user said, but never changes how the agent acts. You wrap your existing model calls with a lightweight SDK and one loop runs. Your agent publishes what happened. Reflexio extracts what to do differently from corrections, failed paths, and good outcomes. and the next run reads those learnings back with nothing retrained.
12:43 It keeps tuning each learning against the sessions it helped and the ones it did not. Retires older lessons when newer conversations contradict them and resolves conflicts in the background so behavior does not drift. It scores responses against the unogmented one so you can tell whether a change actually helped and every learning is auditable. You rewrite, approve, reject or delete it and a rejected one drops from retrieval at once.
13:07 A portable skill lets codeex, claude code, or cursor wire it in, and it runs from fully managed to self-hosted with your own keys, database, or cloud. It is built for teams shipping agents. Start free and let your agents learn. Project number 13. Ponytail makes coding agents write less code. Ponytail is a rule set that makes AI coding agents write the least code that works.
13:31 It solves the overengineering agents fall into reaching for 50 lines and a new dependency when the job needs one line. It works like a lazy senior developer running each task up a ladder and stopping at the first rung that holds. Skip what is not needed yet. Reuse what already lives in the codebase. Prefer the standard library, then a native platform feature, then an already installed dependency, then a single line, and only then the minimum new code.
13:58 validation, error handling, security, and accessibility are never simplified away. You install it as a plugin into Clawed Code, Codeex, Copilot, Gemini, Cursor, and many other agents. Then drive it from chat, set how aggressive it is, review the current diff for bloat, audit the whole repo, or track deferred shortcuts in a ledger. It is open- source.
14:19 It is built for developers running coding agents. Install it and let your agent write less. Project number 14, Hyper Probe. debug production without redeploying or restarting. Hyper probe is an AI native production debugger that reads the live state your logs never capture straight from running code. It solves the slow log redeploy wait cycle where the value that explains a failure was never logged and finding it eats hours in a war room.
14:46 When an incident fires, it picks up the page from pager duty data dog or slack reads the logs and traces to locate the exact file in line. then uses your coding agent to drop a readonly virtual breakpoint there with no redeploy or restart. The probe fires asynchronously on live traffic, captures the exact variable values at that line and disappears.
15:06 So the service never pauses and users notice nothing. From that real evidence, it confirms the root cause instead of guessing, catching silent failures, swallowed exceptions, race conditions, and third party contract drift that never paid you. It stays read only with an audit trail. Runs in your own infrastructure or VPC with PII redacted before capture and supports JavaScript, TypeScript, Java, Python and Ruby with agents like Cursor, Claude Code and Codeex.
15:34 It is built for engineering and on call teams. Try it and work one real incident. Project number 15. Mash Experiential Labs open-source AI gateway one key. Experiential Labs is an open-source AI gateway that puts every model behind one key at the provider's price. It solves the sprawl of juggling separate provider APIs, unpredictable spend, and no clear view of what each agent costs.
15:59 You change the base URL and keep your code, and one OpenAI compatible endpoint reaches hosted providers, your own keys and your own GPUs with new models available on day zero and no markup on routed tokens. An intelligence layer sits on top. It watches your traffic and flags when switching models wins, raises cash hit rate to bring repeated tokens back cheaply, and can fine-tune a model on your own traffic.
16:25 Proven in simulation before it serves and reach through the same endpoint. One console shows every request and dollar by agent, person, model, or day. An admins set per keycaps and model allow lists enforced on every call. You self-host it or use the hosted gateway, and it is written in Rust. It is built for developers and teams running models in production.
16:45 Get an API key and point your agent here. Project number 16, Lightfield AI native CRM that works for you. Lightfield is an AI native CRM that runs the busy work instead of asking your team to log it. It solves the field updates, note-taking, and follow-up drafting that legacy CRM push onto sellers from your calls, emails, and meetings. It builds a living version picture of every customer and agents work off that full context to generate pipeline, prep you before meetings and reopen deals that stalled.
17:18 You ask questions in plain language and get answers cited back to the original conversations. So nothing is a black box and you assemble your own agents from skills, knowledge, and automations wired in over APIs or MCP. It is built for early stage high growth teams that want their CRM to do the work. Try it free and let the agents run. Project number 17, Git Warren.
17:39 Review agent code before it commits. Git Warren is a desktop app for local Git code review that reads an agent's changes before they become a commit. It solves the gap other review tools leave where a review waits for a pull request by which point the work is committed, pushed, and its shape already accepted. When an agent finishes and leaves a work tree full of uncommitted changes, Git Warren finds that work tree wherever it lives on disk and folds its staged, unstaged, and unttracked files into one diff you read and
18:09 comment on right there. It ships an MCP server, so you point Claude Code, Codeex, or any MCP client at it, and the agent gets the same tools the app uses. Opening a review, reading the discussion, replying in a thread, and answering the question you left on a line, all in the review where the answer stays. Machine comments are always marked as such.
18:30 Two agents working at once are kept distinct, and you can edit or delete anything. Nothing leaves your machine, no account, nothing cached, and reviews live in one SQLite file. It runs on Mac OS, Windows, and Linux, and is free and open- source. It is built for developers running coding agents. Download it and read the diff before you push. Project number 18, Brick Forger AI.
18:54 Turn a prompt into buildable bricks. Brick Forger AI is a web tool that turns a text prompt into a custom buildable brick model. It solves the gap between imagining a brick creation and actually having the design, parts, and steps to make it. You describe what you want, a spaceship, a house, a dragon. Pick a build size, and it generates a 3D model you preview with its part count and dominant colors.
19:16 It builds from a real purchasable brick library with a range of slopes and curves for smoother surfaces. And most models come out fully connected and pass a stability checker with organic shapes like animals and plants its strongest work. When you are ready, you download an LDR file and a step-by-step PDF build guide. Then order the bricks and build it for real.
19:38 A free plan gives a few credits each month, and you pay only to download the file. It is built for brick builders and hobbyists. Describe something and forge your first model. Project number 19, GPT6 Astra. OpenAI's frontier model for agentic work. GPT6 Astra is OpenAI's new flagship model built for computer use, coding, and professional work. It targets the gap where earlier models could reason, but struggled to carry out long multi-step tasks reliably inside real software.
20:06 It operates a computer directly, so it fills out forms, updates CRM records, organizes a calendar, researches online, analyzes data, and generates plots, builds and QA checks a website, and installs and troubleshoots software. It produces polished documents, spreadsheets, and presentations that follow your templates and style, pulling only the context that matters, and it stays oriented as a task evolves, filling routine gaps and asking focused questions when an answer could change the outcome.
20:39 In Codeex, it keeps searchable notes across context windows instead of compressing everything into one summary. It is also OpenAI's most aligned model, respecting task boundaries and communicating its limits more accurately. and it advances scientific and mathematical work. It rolls out in chat GPT for plus pro, business and enterprise and through the open AAI API, Azure and Bedrock.
21:01 It is built for professionals and developers running agentic work. Try it in chat GPT or build on the API. Project number 20, Compliance by 12 Labs video compliance review in minutes. Compliance by 12 Labs is a video AI application that reviews your video library against compliance rules you author and control. It solves the manual bottleneck where reviewers watch every frame by hand to catch violence, profanity or brand safety issues, a process that takes weeks.
21:30 It ingests your content and runs it against regional or custom rule packs using multimodal retrieval and contextual reasoning. So, it surfaces not just what happened, but why it may break a rule. Your compliance team forks a regional pack, edits it, tunes thresholds, and publishes a new version release on its own with the rule book living in your tenant rather than the vendor's code.
21:54 And over 40 packs for standards like US, UK, French, Saudi, Brazilian, and Australian bodies ready to start from. Reviewers get a ranked queue of flagged moments with timestamped evidence, accept, reject, and annotate controls and exportable signed reports. It is a fully managed SAS backed by a REST API with role-based access, encryption, and signed URL media delivery. It is built for media, advertising, and broadcast compliance teams. Talk to sales to get your first pack live. Thanks for watching. See you in the next