Featured projects

Publish the output, not the project

Use a clean build folder and filter secrets, repository metadata, dependencies and symlinks out of the upload.

Keep the publishing decision explicit

A draft stays local until the user asks to publish or confirms the public upload.

Keep control of the URL

Save the returned owner token privately so later updates belong to the same owner.

Ground the rewrite

Map the target job to the candidate’s real experience and keep team contributions distinct from individual claims.

Get an independent review

Use an AI from a different company to challenge a draft without being told which model wrote it.

Finish with evidence

Review checker findings, approve changed claims and prepare for the interview questions the resume invites.

Choose the exchange SDK

Map the exchange and runtime to the documented Siebly package instead of inventing a generic API client.

Handle live data properly

Distinguish socket open, subscription acknowledgement, hydration, replay and application readiness.

Make recovery explicit

Add credential validation, rate-limit handling, reconnect logic, durable intent records and duplicate-call protection.

Coordinate a team of agents

Configure agents individually and let them exchange messages while sharing a persistent memory wiki.

Build repeatable workflows

Use runner steps, owner review, hard checks and a record of recent runs.

Connect the working tools

Configure common or per-agent MCP connections and schedule interval, daily, weekly or one-off routines.

Find the right destination

Search destinations and explore their published highlights, categories and activities.

Compare places with context

Read place overviews, selected sections and visiting facts, then discover related places.

Go deeper without starting over

Explore curated collections, page FAQs and documentation through purpose-built readers.

Compare the plans

Inspect pricing, available data, provider scores and plan limits using the published dataset.

Query your requirements

Use DuckDB, jq or Python examples to filter plans by API, budget and other fields.

Keep the source links

Return the selected plan pages and provider documentation alongside the comparison.

Find the relevant facts

Translate job requirements into focused questions about the candidate’s actual work.

Keep claims grounded

Check personal contribution, delivery stage and the source of every number before strengthening a claim.

Prepare for follow-up questions

Audit the draft and rehearse evidence-backed explanations of experience and gaps.

Compare explanations

Keep two or three meaningful diagnoses in play and search for evidence that could weaken each one.

Make the smallest useful revision

Use counterfactual tests and protected author intentions to choose a focused intervention.

Check what changed

Reassess the revision for improvement, regression and remaining uncertainty before another loop.

agentswarm-work

Inspect one agentswarm.space task, reuse an authorized identity when needed, and submit real work evidence within the user’s chosen scope.

cloudgpu-ai-short-drama

Plan staged GPU-rendered shots with open video models, character LoRA and image workflows, then export reusable assets within an approved budget.

cloudgpu-comfyui-images

Run an approved ComfyUI image workflow on rented GPU compute and retrieve images with their workflow metadata.

cloudgpu-discover-compute

Compare GPU capacity and task costs without creating a paid instance, then prepare a proposal for user approval.

cloudgpu-lora-finetune

Plan a bounded LoRA or QLoRA training run on rented GPUs and export adapters, checkpoints and metrics.

cloudgpu-open-model-inference

Run an approved open-model inference or evaluation workload with protected serving and explicit output retrieval.

Plan the campaign

Build useful briefs, messages and measurement plans from approved offers and audience facts.

Design the customer journey

Specify triggers, suppression rules, branches and exits for lifecycle automation and audience segments.

Match the industry

Use dedicated guidance for ecommerce, SaaS, education, events and eleven other industries while keeping their data and review boundaries clear.

Create authenticated checkout

Connect a signed-in user to a Polar subscription checkout and a payment-pending success screen.

Handle payment events

Verify webhook signatures, deduplicate deliveries, allowlist products and guard against stale entitlement updates.

Check launch readiness

Review production configuration, access checks and billing evidence with PASS, FAIL and NOT RUN outcomes.

Show the actual interface

Use readable real screenshots, deliberate close-ups and labels that follow the action.

Keep the scene editable

Separate artwork from SVG/HTML foregrounds and procedural 3D so revisions stay precise.

Capture and inspect the export

Use exact-frame local capture and encoded MP4 checks before handing over the video.

Write tests in plain language

Describe a user journey and agent-qa tests it on web, Android or iOS.

Keep the evidence

Inspect each failed step beside its screenshot and the recorded run evidence in the dashboard.

Remember what works

memidex keeps facts and working procedures as reviewable files in your repository, and agent-qa validates lessons from each run.

Work from your coding agent

Author tests, triage failures, read run evidence and debug fixes through the agent-qa MCP tools and skills.

Find roles

Search with German and English titles across sources chosen for your field, drop off-target titles and save a ranked shortlist.

Check a posting

See the German requirement with the quoted line, the salary against 2026 Blue Card thresholds, and agency and contract terms.

Apply and interview

Prepare a German Lebenslauf and Anschreiben or an English CV and letter, checked against your CV, then practise mock interviews with scored feedback.

Track applications

Log application news, draft follow-ups and see response rates on an offline dashboard.

Draft and schedule posts

Review the draft first, then schedule it, publish it or queue it in the next free slot.

Publish in bulk

Upload media from files or URLs, manage the media library and batch-create posts.

Manage channels and the inbox

Check connected channels and their health, and answer DMs, reviews and comments with your yes on each reply.

Measure and plan

Pull engagement analytics, check plan limits and usage, and follow each platform’s post types, media rules and character limits.

Ask by voice or text

Tap the voice button or type the task you need done.

Work from your details

Empacer prepares the work from your business details, with photos, price lists or invoices attached when you need them.

Approve every result

Each result arrives as a card you can edit, send back for changes, decline or approve. By default, nothing is sent until you approve it.

Approve the plan first

Sign off on the storyboard before anything is drawn, and point at any page to say what to change.

Narration and voice

The agent writes the script and voices it with ElevenLabs, or with Volcengine TTS for Chinese.

Pages in your style

Every page is drawn in a built-in style (warm-story, study, chalkboard, whiteboard or tech-dark) and checked with lint and screenshots.

A finished film

A final check on the full video, then a 1080p MP4 with subtitles, with cost and time shown at every step.

inkroom-prompts

Search with one to three keyword groups, review up to three prompts with their example images, and optimize your own idea into a prompt when nothing matches. It also answers questions about public InkRoom models, prices and capabilities.

inkroom-prompt-reverse

Reconstruct Chinese and English image-generation prompts from one image through an existing InkRoom MCP connection, with a cost preview before analysis.

Set up a mail server

Run lfr setup with your mail hostname, add a domain and publish exactly the DNS records lfr prints; lfr doctor checks them.

Run mailboxes

Create addresses, aliases and forwarding, search and answer mail, and train the spam filter as you move messages.

Publish newsletters

Run lists with double opt-in and stored consent proof, Markdown issues checked before sending, one-click unsubscribe and a website for each list.

Diverge on ideas together, then converge on priorities and next steps in a session document.

Examine the claims, evidence, assumptions, gaps and fallacies in a memo or draft, each anchored to quoted text.

Compare candidate CVs with a job description on a shared 100-point rubric with traceable evidence.

Turn business questions into defensible analysis, validated models and decision-ready reports.

Work from real user evidence through empathize, define, ideate, prototype and test.

Take a bare idea to a polished prose manuscript with a team of named specialist agents.

Run citation-backed market scans, sizing, competitor analysis, demand signals and trend analysis.

Turn a recurring failure into a precise problem statement, verified root causes, traced solutions and a rollout plan.

Prioritize the backlog, write PRDs and user stories, set metrics and OKRs, and plan launches in a product workspace.

Keep charters, schedules, estimates, risk registers, status reports and scope changes in a project workspace.

Plan sprints, run daily pulses and retrospectives, track impediments and spot process anti-patterns.

Explore a topic or a document in dialogue with Gadfly, a Socratic questioning partner.

Take a strategic question to a defended, board-ready recommendation with a team of named specialist advisers.

Choose tools for live search, page data, visual captures and website health checks through Geekflare MCP or its REST API.

Inspect source without running the skill

Combine pattern detection with Python and Bash dataflow analysis. Review the source as data before installing its scripts or hooks.

Keep evidence with the findings

Export JSON or SARIF, retain the exact source revision and scanner version, and inspect errors and coverage gaps alongside findings.

Choose what leaves the machine

Local analysis can run in an isolated environment with network access disabled. Optional hosted LLM, AI Defense, OSV and VirusTotal routes can send data when enabled.

Draft distinct social posts, review account timing, and approve the exact content and schedule before publication.

Connect a publishing destination, explore article topics, and prepare blog drafts for human review.

Research a product category, prepare an approved Toolfound submission, and track owner verification and launch status.

Create a local Excel budget workbook with household, sinking-fund, and side-hustle inputs.

Review pasted listing copy for length, repeated terms, risky wording, and keyword coverage using a local Python helper.

Transpose chord symbols locally or use Clef MCP tools for piano chords, guitar-tab notes, and music practice.

One workflow: extra-caffeine-workflow

Choose the duration and display behavior, verify awake state, and reconcile timeouts before retrying.

Compare compute and storage billing, explore query and dataset costs, and review allocation options.

Trace contributors and timing, check usage and operational evidence, and separate supported explanations from assumptions.

Draft custom cost axes, preview leftovers and rule overlap, then publish approved allocation rules.

Choose the Costory tools and detailed skills for spend queries, investigations, budgets, dashboards, reports and alerts.

Create author profiles and generate articles through Appie’s hosted MCP service.

Supply knowledge sources, track training and chat with a bot through Appie.

Create short links and retrieve their click statistics from your agent.

Measure the overhead

Inspect always-loaded memory and governance files before choosing a maintenance pass.

Review the changes

Use snapshots and rollback tools to inspect and recover transaction changes.

Check structured tokens

Detect drops in tracked tokens such as paths, numbers and commands. Review prose meaning separately.