Agent PlatformConceptAgentry
Build, run and govern multi-agent workflows.
An orchestration platform for multi-agent AI workflows: a visual builder, full run traces, human approvals, evaluations and cost control.
- Category
- AI · Agent Orchestration
- Designed for
- Web
Why Agentry exists
The problem
Teams build AI agents as scripts and demos, then cannot see what they did, control what they cost or stop them doing something they should not.
The product
Agentry gives multi-agent workflows a visual builder, full traces, human approvals, evaluations that gate deploys and budget controls.
Built for: Operations and platform teams putting AI agents into real business processes.
- Visual builder for orchestrator and specialist agents
- Tools connected through the Model Context Protocol
- Step-by-step run traces with cost and latency
- Human approval queue for high-impact actions
- Evaluation suites that gate every deployment
- Model routing and per-workflow budgets
Walkthrough
Agents You Can See Inside
Designed for the engineer on call: every step, tool call, token and decision visible and replayable.
- 01
Workflow Builder
An orchestrator routes work to specialist agents with guardrail and approval steps, shown with live success, cost and eval scores.
- 02
Run Trace
Each step on a timeline with tokens, cost and tool calls, plus the agent's reasoning summary and the exact tool payload.
- 03
Approvals Queue
Actions over a limit wait for a person, with the evidence and a link to the full trace.
- 04
Cost, Latency & Evals
Spend by workflow, model routing mix and eval results for each version, with regressions blocking release.
Every screen
The full experience.
What makes it different
Governance Is Part Of The Runtime
Approvals, budgets, evaluations and tool permissions are enforced by the platform, not left to each workflow's prompt.
Versioned Workflows
Every workflow change is a version with a diff and an eval score.
MCP-Native Tools
Agents reach business systems through scoped Model Context Protocol servers.
Policy Enforcement
Approval thresholds, blocked actions and data rules applied at run time.
Model Routing
Small models for simple steps, larger models only where reasoning is needed.
How it’s built
The technology stack.
The architecture we would build this on today, and why each piece earns its place.
- Web app
Next.js, React, TypeScript, Tailwind CSS
Review queues and dashboards that stay fast with thousands of items
- AI services
Python (FastAPI) · Claude via the Anthropic API with tool use
Strong reasoning and structured tool calls, isolated from the core API
- Tools & integrations
Model Context Protocol (MCP) servers
One standard way for agents to reach business systems, with scoped permissions
- Durable workflows
Temporal
Long-running, multi-step AI work that survives restarts and waits for approvals
- Knowledge & memory
PostgreSQL + pgvector
Retrieval over approved sources without a separate vector database
- Agent protocols
MCP for tools · A2A for agent-to-agent
Open standards so agents and tools are not locked to one vendor
- Evaluation & tracing
Versioned eval suites, OpenTelemetry traces, cost tracking
Every prompt or model change is tested before release and every run is explainable
More inside
Build, Observe, Govern.
The lifecycle of an AI workflow in production.
Agent Library
Reusable agents with their instructions, tools and test cases.
Memory Controls
What agents may remember, for how long and from which sources.
What comes next
On the roadmap.
What a second release would add once the first one is in real users’ hands.
- 01Agent-to-agent (A2A) interoperability
- 02Canary releases for workflow versions
- 03Red-team scenario library
- 04Self-hosted deployment option
- 05Per-customer budget limits
Related
Services and industries behind this work.
Concept project
Need Something Like This?
This is concept work: the product design, screens and architecture we would start from. If your business needs AI agents in production with proper controls, we can scope a first release with you, usually within a week.
Next project
Receptia
An AI receptionist that answers, books and knows when to escalate.
