klaudijaDedicated agents
Define instructions, models, skills, environments, and agent versions for a specific role.
SMART MIDDLEWARE FOR YOUR AI PRODUCTS
Claude runs the agents.
You run the business.
Klaudija connects the two: your platforms, customer identities, plans, conversations, and company expertise, managed from one place.
Under the hoodTHE BUSINESS LAYER ABOVE THE API
Why use Claude
with Klaudija?
Claude provides the AI runtime. Klaudija adds the product-level identity, ownership, spending, and operating workflows around it. Here is how they fit together.
| Your product needs | Claude Platform provides | Klaudija adds |
|---|---|---|
| Monthly budgets per customer | A cost budget for each Managed Agents session.Session budgets | A monthly allowance per business user or team across their conversations, with billing dates, plan overrides, extra credits, and a customer-facing usage API. |
| User-owned conversations | Stateful sessions, persisted events, and session lifecycle APIs.Session operations | Map your authenticated business identity to its sessions. Apply ownership checks to job status, results, cancellation, files, and history. |
| Several customer-facing platforms | Workspaces, API keys, members, and workspace permissions.Workspaces | A product-level platform catalogue, origin resolution, backend user provisioning, default agents, user overrides, and platform-scoped operator roles. |
| Context that follows the customer | Persistent memory stores that applications can attach to sessions.Agent memory | Provision and reuse memory by business identity, provide operator curation, and retain logical conversation history across session rollover. |
| Cost visibility in business terms | Usage and cost reporting by dimensions such as workspace, API key, and model.Usage and Cost API | Attribute recorded usage to your users, platform sources, jobs, and sessions. Inspect plan state, credits, and usage exports in one operator workflow. |
| Company expertise in daily operation | Agent configurations, tools, skills, and managed execution.Managed Agents | Import desktop bundles, review compatibility, preview AI-assisted conversion, maintain versioned assets, and assign agents to the clients who need them. |
Claude also offers per-user monthly limits in its Claude Code workspace. Klaudija’s plans and allowances apply to the business identities in your own products. Its budget checks use recorded usage; see how allowances work.
Comparison reviewed 5 September 2026 against the linked Claude documentation and Klaudija source. Klaudija implements application-level controls; provider workspace isolation and your product’s entitlement design remain separate configuration decisions.
01 / AGENTS
An intelligent team, built around your work.
Configure the specialists your business needs, then give them a clear way to work together.
Multi-agent coordination
Connect a coordinator with specialist subagents and configure their models and tools.
Direct conversation
Keep everyday interactions moving with direct chat, delegating complex work to managed agents.
Tools & connected systems
Equip agents with configured MCP connections, web access, file tools, and scoped credentials.
AI agent review
Request a critique or adapted system prompt, inspect the proposal, and save a reviewed agent version.
Controlled publishing
Version agent configurations, inspect dependencies, and maintain an operational change history.
Assignment visibility
See the users and platforms that use an agent, and inspect the requirements of its attached MODs.
Tool policies
Choose enabled shell, file, and web tools, and retain configured MCP tools and permission policies.
02 / KNOWLEDGE
Your expertise becomes a working asset.
Build on the knowledge that makes your company yours. Keep it maintainable, reusable, and attached to the right work.
Private skill bundles
Package instructions, reference material, templates, and executable resources as reusable MODs.
Versioned knowledge
Edit assets, inspect previous versions, compare changes, and restore a prior skill version.
Import desktop skills
Bring ZIP or .skill bundles authored in Claude or Codex, then review their tools, paths, and runtime assumptions.
AI-assisted skill review
Check compatibility, preview AI conversion across supporting text files, and inspect diffs before publishing.
Templates, scripts & assets
Inspect bundled files, update individual assets, and publish a new version without rebuilding your expertise.
Credential review
Find likely embedded secrets and review a proposal to replace them with named Vault-backed references.
Fleet compatibility review
Review agents and custom MODs together, with version-specific compatible, warning, and blocked findings.
Persistent user memory
Carry useful context between conversations with user-associated memory stores.
Memory curation
Browse stores and file metadata, inspect and edit useful context, and remove obsolete memory files.
Durable assets
Keep skill files and generated documents in platform storage alongside provider references.
03 / CONTROL
Your rules have a place to live.
Operate across platforms and clients with spending visibility, user plans, and permissions designed into the workflow.
Platform management
Manage tenant platforms, allowed origins, dedicated agents, and tenant-scoped administration.
Identity mapping
Connect external identities to canonical users, including shared team budget identities.
Plans & credits
Assign allowances, billing periods, plan overrides, and additional credits to the right account.
Usage & cost visibility
Inspect model usage, token categories, spend attribution, and configured rate cards.
Session lifecycle controls
Configure pre-turn rollover thresholds for model requests, elapsed time, and recorded cost while preserving logical history.
Scoped permissions
Assign owner, admin, AI manager, platform manager, platform viewer, or tester access.
Credential management
Keep agent secrets in scoped Vault credentials with host restrictions.
Operational traceability
Review administrative changes, job events, audit records, and errors from one console.
04 / EXECUTION
From request to a result you can use.
Run the work your product promises, with a durable trail from the initial message to the final document.
Asynchronous jobs
Submit work, receive a job ID, and follow completion without a long-held client request.
Queued follow-ups
Keep pending messages in a durable queue while a turn runs, with client-driven dispatch when ready.
Live progress
Deliver job and session events to the interface through Supabase Realtime.
Documents in, deliverables out
Attach supported files and retrieve generated deliverables, including an archive when the agent produced one.
Session continuity
Resume conversations, cancel active work, and inspect archived history.
Recovery & reconciliation
Recover event history and reconcile stalled work with operator-visible recovery actions.
Declared runtime requirements
Record packages, network hosts, and named secrets for a MOD, then prepare the configured environment.
Health & latency telemetry
Inspect operation reliability, timing percentiles, slow turns, errors, and recovery incidence.
Spending exports
Analyze recorded costs across users, sources, models, and sessions, then export usage as CSV.
Operator playground
Test up to four parallel chat panes with conversation, tool, and internal event views.
05 / VOICE
Speak. Practice. Get better.
Bring natural voice into the experience, and give your people a place to rehearse real conversations.
Voice conversations
Talk to Klaudija and keep the conversation connected to your business context.
Configurable scenarios
Create public briefs, private personas, goals, and evaluation guidance for each training exercise.
Personality & difficulty
Shape anxiety, scepticism, mood, talkativeness, literacy, cost sensitivity, and difficulty.
Voice & pacing
Configure how the persona sounds and the pace of the interaction.
Coaching reports
Review category scores, strengths, improvement areas, and moments worth discussing.
Transcripts & replayable practice
Revisit a conversation and repeat a scenario with the same settings.
06 / DEVELOPER
Built to fit into your product.
Keep ownership of your customer experience while Klaudija handles the intelligence behind it.
JWT-based user access
Authenticate customer-facing calls with scoped end-user tokens.
Server-side platform keys
Provision identities and manage platform users from your trusted backend.
Async API & file delivery
Integrate job submission, status polling, result download, and user usage endpoints.
Operator APIs
Manage skills, agents, sessions, platforms, users, and settings programmatically.
Integration documentation
Start with request examples, authentication guidance, response shapes, and error handling.
Claude-powered agents
Use Claude’s managed execution through Klaudija’s platform, identity, and operational layer.
A PLATFORM WITH A POINT OF VIEW
Your business context
is the real advantage.
Models provide intelligence. Your workflows, knowledge, customer relationships, and operating rules make it valuable. Klaudija gives those assets a home.
YOUR NEXT CHAPTER
Give your business
a mind of its own.
Your people. Your knowledge. Your rules.
Meet the AI that brings them together.