Features | CCD by DENDAT

The Agent

A persistent AI engineer, not a chatbot

Give it outcomes instead of instructions. It breaks the goal into steps, executes them on your node, and streams every file change and command as it happens.

Persistent sessions

Every conversation is a session tab that survives restarts. Leave, come back, and continue exactly where you left off. Everything older lives in History.

Planning next step…

npm run build ✓

Recalled 14 project facts

Outcomes, not commands

Say 'add auth to the Flask app in ~/projects/shop and make the tests pass.' The agent plans the steps and runs them for you.

Long-term memory

Hierarchical and semantic memory recall your stack, decisions, and past fixes across sessions, so you never re-explain the project.

Fixes its own errors

When a build breaks or a test fails, it reads the output and corrects course instead of stopping and waiting.

Verified progress

Work is checkpointed to git with the exact commands it ran, so every step is auditable and replayable.

8 specialized modes

Developer, Security, Research, Architect, DevOps, Teacher, Debugger, and Reviewer. Each tunes focus and tool permissions.

Approvals and guardrails

Sensitive tools like shell, writes, and dispatch pause for your approval depending on mode. Read-only modes can never modify anything.

MULTI node runs

Fan one task out to every online node, one tab per node plus an overview grid. Folder-locked tasks collapse to a single node for safety.

Eight modes, one agent

👨‍💻

DeveloperFull-stack code, production-ready, with tests.exec requires approval

🔒

Cyber / SecurityAudits, vulnerability scanning, threat analysis.security-scoped tools

📊

Research & AnalysisData analysis, research synthesis, docs.read-only

🏗️

ArchitectSystem design, scalability, tech evaluation.exec & writes require approval

⚙️

DevOpsCI/CD, infra automation, deployments.infra tools require approval

📚

TeacherExplanations, tutorials, concept breakdowns.read-only

🐛

DebuggerRoot-cause analysis, logs, surgical fixes.writes require approval

🔍

Code ReviewerQuality, security, and performance review.read-only

The Fleet

Your own isolated Linux machines

A node is a dedicated machine assigned to your company. Your code lives there, work runs there, and nobody else's workloads share it.

Dedicated nodes

Not a shared multi-tenant runtime. 5 nodes on Team, 12 on Scale, a full fleet on Enterprise, each running its own persistent agent.

node-07 · your company● online

node-12 · your company● online

Live node health

Every node shows live status: online and ready, or offline with tasks auto-routed elsewhere. Load and capabilities are visible per node.

Dispatch, broadcast, chain

Send a task to one node, broadcast the same task to many at once, or chain agents so one node output feeds the next.

GPU nodes

GPUs are detected automatically by type, VRAM, and CUDA capability, unlocking GPU task types like ML training and rendering. Quality adapts to the hardware.

Self-healing

A health monitor watches every node and auto-recovers wedged processes without you noticing.

Folder locks

Tasks scoped to a project folder are serialized to a single node, so concurrent agents can never corrupt a project.

Ship it

From localhost to a public URL, instantly

No deploy scripts, no DNS, no certificates. The moment an app builds, it is live on a managed link you can share.

Instant HTTPS tunnels

Every app gets a managed public https URL the moment it starts listening on a port. Zero config, fully managed.

app-saas.dendat.app● live

App Store

Deploy, monitor, and restart every project from one dashboard with live logs and health checks.

Projects and workspace

A project manager and file workspace on every node, with full history of your past sessions.

Publish apps

Turn a working build into a shareable public app page in a click.

Docs generator

Generate real documentation for your project straight from the workspace.

Git built in

Checkpoints, diffs, and a git UI so every change the agent makes is tracked and reversible.

Trust

Hard isolation and real guardrails

Security is not a setting you turn on. It is how the platform is built, from where your code lives to how every request reaches the app.

Dedicated, not multi-tenant

Your projects live on machines assigned to your company, walled off from every other tenant on the platform.

Roles and admin mode

Admin, developer, and viewer roles. Platform administration needs an explicit admin mode, so admins can never trigger destructive actions by accident.

PBKDF2 authentication

Passwords are stored as salted PBKDF2-HMAC-SHA256 hashes at 260,000 iterations, never in plaintext. Sign in by browser session or company X-API-Key.

Sandboxed execution

Agent commands can run inside an isolation sandbox on the node when available.

Deep redaction

Logs and transcripts are scrubbed of API keys, tokens, and credential-shaped strings before they are stored or shown.

Encrypted BYOK

Your provider keys are encrypted at rest with a key held outside the database.

Scanner shield

Every request is fronted by a shield that blocks bot and vulnerability-scanner traffic, probes for .env, .git, and admin panels, before it reaches the app.

GET /.env403 blocked

GET /wp-admin403 blocked

Audit and isolation

Full audit logging across your company, on top of hard tenant boundaries between every organization.

Control

Predictable cost, open access

Bring your own AI keys, meter usage across your company, and drive everything from a real API when you want to.

Flexible plans and seats

Team, Scale, and Enterprise plans. Add seats as your team grows.

Annual billing

Pay for ten months on annual billing, handled through Stripe.

Usage metering

See usage and cost analytics across your whole company.

Bring your own keys

Attach your own DeepSeek or Groq key and the agent uses your key first, so your runs keep going on your own quota. Encrypted at rest.

DeepSeek · your key● routing first

Full API access

Start agent tasks from a script with your X-API-Key, against a live, self-documenting API reference at /api-docs.

Smart model routing

A model router picks cost-effective models per task, so you are not overpaying for simple work.

Beyond code

Creative studios on the same fleet

The agent does more than ship software. These studios run on the same dedicated nodes.

Builder, the game studio

Describe a game and the agent scaffolds it with a full dev loop: live reload while it builds and an on-screen error overlay so problems surface immediately.

live reload✓ hot

Worlds generator

Composes terrain, environment, and detail layers into explorable 3D worlds rendered in the browser, with thumbnails saved to your workspace.

The Arcade

A public gallery of playable simulations that adapt quality to the GPU they detect, so the same page runs from a laptop to a workstation.

Residency and DenVoice

Turn a node into a voiced companion using DenVoice, our own local formant synthesizer. No cloud text-to-speech, so conversations never leave your fleet.

The long tail

And a couple hundred more

The headline features are the tip of it. A sample of what else ships on the fleet.

Reasoning engineHierarchical plannerSelf-reflectionCreative solverMulti-agent orchestrationTree-sitter code indexSemantic memoryPolicy engineInput validationCost analyticsModel routerRender brokerllvmpipe fallbackGPU worker watchdogCloudflare tunnelsLive reloadError overlaySession historyAgent chainPipeline chainingDispatchBroadcastFolder locksApprovalsX-API-Key authQuotasRate limitsDeep redactionFirejail sandboxScanner shieldPBKDF2 authAdmin modeAudit logsTenant isolationBYOK routingDeepSeekGroqAnnual billingSeatsUsage meteringProject managerWorkspaceDocs generatorPublish appsApp storeHealth monitorGPU detectionWorldsModelforgeGame studioDenUniverseResidencyDenVoiceAPI reference

Everything an engineering team does, on every node