Skip to content

    Products from Algen R&D

    Products for theagent lifecycle.

    Developed by Algen's product and applied research team: build and run agents, inspect packages before adoption, and observe, evaluate, and govern agents across the frameworks you already use.

    Use what fits. The products are designed to work independently as well as together.

    How the ecosystem fits

    01

    Build & run

    Agent Runtime + Agent Studio

    Write agents in code with Runtime, or design, test, and ship them visually in Studio, on the same projects.

    Explore
    02

    Discover & distribute

    Agent Hub

    Find versioned agents with AgentStripes ratings, declared permissions, and clear delivery paths.

    Explore
    03

    Observe & govern

    Traccia · our flagship product

    Observe, evaluate, govern, and prove what production agents do, across models and frameworks.

    Explore

    Algen Agent Runtime

    Open source · Apache-2.0

    Algen Agent Runtime: Build governed agents.Own the runtime.

    A typed, provider-neutral Python runtime for governed agents and multi-agent workflows. Approvals, budgets, and policies run in the execution path, so the model can't talk its way past them.

    Explore on GitHubRead the docs

    Pre-release, hardened for a single node. View on PyPI

    See it in a minute
    $ python -m pip install --pre algen-agent-runtime
$ python -m examples.quickstart_approval.app

    No API key and no network. The workflow proposes a change and pauses for your approval. Approve it and the change is applied once, under an idempotency key; reject it and nothing is written.

    1. Governed tools

      01

      Tools declare their side effects. Writes wait for a human approval in the execution path, not the prompt, and run under an idempotency key.

      Read the post

    2. Durable runs

      02

      Runs, workflows, and approvals are checkpointed and survive a restart. An interrupted write is reconciled, never blindly retried.

      Read the post

    3. Any model

      03

      Route across hosted and local providers with fallbacks, run budgets, and per-account credit metering on every call.

      Read the post

    4. Workflows

      04

      Declare DAGs of agents, handlers, and joins with typed inputs, bounded repairs, and child workflows. The runtime schedules and recovers them.

    • Python library or REST/SSE service
    • OpenTelemetry built in
    • Offline tests with a mock model

    Algen Agent Studio

    Visual agent workbench

    Algen Agent Studio: From idea to agent,on one canvas.

    Design, test, build, and ship governed agents in one workspace. Every Studio project is an Algen Agent Runtime project, so you can export it and run it with Runtime yourself.

    Read the user guide

    Invite only during the preview. Already have access? Sign in

    Built-in assistant

    azAI builds with you, and asks first.

    • Plans a multi-part build for you to approve, then builds it.
    • Edits the draft, with Keep and Undo on every change.
    • Asks before anything that costs credits or leaves Studio, and stops at the limit you set.

    1. Design

      01

      Describe agents in plain language, choose their models, chain them into multi-agent workflows on a canvas, and connect tools and data.

    2. Test

      02

      Chat in the Playground, inspect every step of a run, and lock behavior in with evaluation suites before you promote a change.

    3. Build

      03

      Package custom code into an image that is scanned and signed in an isolated builder, with an SBOM and provenance.

    4. Ship

      04

      Deploy to Docker, Podman, or AWS from a plan you review, download a bundle to run yourself, or publish to Agent Hub.

    • Organizations, roles, and invitations
    • Credentials encrypted at rest
    • Credit-metered runs

    Traccia

    AI agent control plane

    Flagship product

    Traccia: Let your agents do more.Without losing control.

    An OpenTelemetry-native control plane to observe, evaluate, govern, and prove what production AI agents do, across models, frameworks, and clouds.

    Explore TracciaRead the docs

    Free plan with no card, and a 14-day trial of every feature. See pricing

    Instrument in two lines
    $ pip install traccia

from traccia import init, observe
init()

    Decorate your functions with @observe() and traces start arriving. The Python and TypeScript SDKs are open source and export to any OpenTelemetry backend, even without a Traccia account.

    1. Observe

      01

      Traces with execution timelines, per-agent dashboards, and cost attributed by agent, model, and trace. Claude Code, Cursor, and Copilot can send telemetry too.

    2. Evaluate

      02

      Versioned prompts with deploy labels, datasets, LLM-judge and code scorers, and experiments that can gate CI.

    3. Govern

      03

      Policies that deny, reshape, or hold a call, from 18 templates. Simulate one on past traffic, and send refunds and purchases to a human approval inbox.

      Try it in Heist the Agent

    4. Prove

      04

      A Compliance Hub for AI systems, reviews, incidents, evidence, and audit logs, with EU AI Act and HIPAA modules on the Enterprise plan.

    • Gateway for OpenAI, Anthropic, and Gemini traffic
    • LangChain, CrewAI, OpenAI Agents SDK, and more
    • SOC 2 Type II in progress

    Algen Agent Hub

    Governed agent marketplace

    Algen Agent Hub: Find your next agent.Know what you're adopting.

    Discover, evaluate, and adopt AI agents as versioned releases, with AgentStripes ratings, declared permissions, and clear delivery paths. Agent Hub is framework-neutral and never runs a package for you.

    Explore Agent HubCheck your agent

    Browsing needs no account. Publishing and adopting are free.

    AgentStripes on every listing
    1. 1
    2. 2
    3. 3
    4. 4
    5. 5
    6. 6
    7. 7

    No stripes yet

    1–2 read from the code3–7 need evidence from real runs

    Each release is rated from its code, and every rating cites the code it rests on. Two stripes is the bar for a listing. Stripes describe maturity; they aren't a safety guarantee. How stripes work

    1. Discover

      01

      Search and filter by category, framework, and minimum stripes. Every listing shows its capabilities, its boundaries, and the release it describes.

    2. Inspect

      02

      See declared permissions, data classes, outbound hosts, and runtime secrets before you adopt, with reviews tied to the version they rate.

    3. Adopt

      03

      Download a digest-pinned package and verify its SHA-256, run a source container, or request access to a publisher-hosted endpoint.

    4. Publish

      04

      Upload an immutable package. It is quarantined, scanned, needs two stripes, and is reviewed before it goes live. Studio can publish for you.

    • Algen Runtime, LangGraph, CrewAI, OpenAI Agents SDK, and containers
    • Ratings only from people who adopted the version
    • Free to publish and adopt

    Need a product—or a system built around it?

    Explore the public products directly, or work with Algen Engineering to connect AI to your workflows and existing systems.

    Explore Engineering

    Product details and public destinations reviewed on November 7, 2026.