AgentPlatThe AgentPlat library
Ideas and practical scenarios for leaders building AI products with governed autonomy.
Start here · AGI readiness
Why AI products need a durable purpose, clear authority and accountable work as agents become more capable than their human operators.
Read the guidePrepare authority and evidence boundaries for increasingly capable agents.
AGI readiness
Why AI products need a durable purpose, clear authority and accountable work as agents become more capable than their human operators.
Read articleUnderstand purpose, inceptions, signals, limits and the bounded execution cycle.
Governed autonomy
Give an AI product a continuing responsibility its customers can understand, guide and evaluate.
Read articleGoverned autonomy
How an agent turns a customer’s idea, request or problem into considered work toward a purpose.
Read articleGoverned autonomy
How signals help AI products notice changing conditions and decide when a customer’s goal needs attention.
Read articleGoverned autonomy
Help agents interpret customer expectations while keeping business rules and permissions clear.
Read articleGoverned autonomy
Make the scope of an autonomous product explicit: what it can do, what requires approval and who can change the boundary.
Read articleGoverned autonomy
Give customers an understandable resource allowance for ongoing agent work, including requests that are still in progress.
Read articleGoverned autonomy
Turn a broad customer goal into manageable work with a clear deliverable, progress and review.
Read articleGoverned autonomy
Explain what an agent actually achieved, what the evidence supports and what still needs attention.
Read articleGoverned autonomy
Keep customers in control of an agent’s purpose, responsibilities and decisions as autonomy grows.
Read articleGoverned autonomy
Preserve customer responsibilities and constraints as agents delegate work and adopt stronger models.
Read articleGoverned autonomy
Expand an AI product’s autonomy from demonstrated behavior and clear customer value.
Read articleGoverned autonomy
Choose between immediate task-based assistance and continuing autonomous responsibility within the same product.
Read articleApply the concepts to support, operations and evidence-led research.
Practical purpose workflows
See how a support product can investigate requests, prepare evidenced answers and retain human control over commitments.
Read articlePractical purpose workflows
Build proactive operations assistance that notices changes, investigates them and helps people choose a response.
Read articlePractical purpose workflows
Create a research product that preserves its sources, supports revisable conclusions and makes uncertainty understandable.
Read articleChoose a work model and plan your integration.
Foundations
Choose whether AgentPlat fits your TypeScript application by examining shared work, review requirements, deployment ownership and simpler execution paths.
Read articleFoundations
Understand the responsibilities of Agent Rooms, Collective Runtime, Agent Mesh, Morphogenesis and the adapters that connect AgentPlat to your infrastructure.
Read articleFoundations
Plan an Agent Rooms deployment around verified identity, persistent storage, provider adapters and operational checks, starting from the self-hosted reference app.
Read articleFoundations
Compare Room collaboration, coordination workers and durable workflow engines, including the role of the optional AgentPlat Rooms Temporal adapter.
Read articleFoundations
Start with AgentPlat’s deterministic local execution, then connect an explicit model adapter while keeping credentials and persistence expectations clear.
Read articleConnect providers, tools and external systems through explicit application boundaries.
Integrations & operations
Connect an MCP client to your own AgentPlat Rooms with scoped authorization, audit records and durable idempotency for mutation tools.
Read articleIntegrations & operations
Introduce agent tools incrementally by separating reads from external effects, recording scope and testing provider paths against application-owned controls.
Read articleIntegrations & operations
Separate application-owned uploads and file retention from Agent Room knowledge revisions, authorization and provenance when connecting documents to agents.
Read articleIntegrations & operations
Treat AgentPlat topology, adapters, policy, recovery budgets and evidence settings as versioned operational inputs when reviewing or reproducing a deployment.
Read articleSee how optional model, assessment and work-management providers connect to AgentPlat.
External providers
Understand the optional AgentPlat Asana adapter, stable external task identity and retry handling while keeping human contribution state authoritative in Rooms.
Read articleExternal providers
Learn how AgentPlat's optional Jev assessor can add typed assessment signals while your application keeps its rubric, policy and approval authority.
Read articleExternal providers
Connect OpenAI or a compatible endpoint through an explicit AgentPlat model adapter, and keep credentials, model selection and inference controls in your application.
Read articleExternal providers
See how AgentPlat’s native Anthropic Messages adapter connects a Claude model while your application controls credentials, prompts and protected actions.
Read articleExternal providers
Choose between AgentPlat’s native Gemini GenerateContent adapter and its OpenAI-compatible route based on the endpoint and model features your application needs.
Read articleConnect people, tasks and versioned work in a shared Room.
Agent Rooms
Learn how persistent human-agent collaboration connects tasks, versioned artifacts and human decisions in an AgentPlat Agent Room.
Read articleAgent Rooms
Understand Agent Room lifecycle transitions, read-only archives, subroom completion constraints and the records that explain progress across a shared objective.
Read articleAgent Rooms
Learn how typed AgentPlat Handoffs transfer context and ownership, and how progressive plans materialize eligible work without duplicating logical steps.
Read articleAgent Rooms
Understand durable human contribution requests, scoped knowledge bundles and external work-management projections while keeping Agent Rooms authoritative.
Read articleAgent Rooms
Understand revisioned execution sessions, durable intervention requests and provider checkpoints for controlling an active Agent Room run.
Read articleAgent Rooms
Learn how to integrate the Agent Rooms HTTP API with injected services, authenticated tenant context, response envelopes and revision-aware client behavior.
Read articleAgent Rooms
Understand Agent Room LiveView, composite cursors and server-sent events so your interface can reconnect without treating a stream as authoritative state.
Read articleAgent Rooms
Plan coordinated Agent Rooms schema upgrades, including migration order, operational projections, worker rollout and restore checks for PostgreSQL deployments.
Read articleAgent Rooms
Understand stable agent identities, published revisions and runtime-profile binding so an Agent Room run can be traced to the exact agent definition used.
Read articleAgent Rooms
Understand how automatic Room coordination commits a message, domain event and initial work item together, and what that atomic boundary does not cover.
Read articleAgent Rooms
Learn the Agent Room routing order, how membership affects eligibility and why selecting a participant remains separate from authorizing a task or effect.
Read articleAgent Rooms
Understand how Room task context selects messages, artifacts and memory, and how provenance helps explain which information an agent used for a run.
Read articleCoordinate independent peers with explicit evidence and membership.
Agent Mesh
Learn when Agent Mesh is useful, how it coordinates independent runtimes and which transport, membership and effect-control responsibilities remain with your host.
Read articleAgent Mesh
Understand bounded Mesh envelopes, sender and scope validation, membership epochs and compatibility checks before peer messages influence agent coordination.
Read articleAgent Mesh
Explore causal anti-entropy, predecessor evidence and readiness gates in Agent Mesh, and why synchronized state does not authorize an agent action.
Read articleAgent Mesh
Learn how Agent Mesh uses governed membership epochs for enrollment, replacement and key rotation, and why old credentials do not establish current authority.
Read articleAgent Mesh
Understand bounded active and reserve peer views, policy-governed adaptation and how to evaluate sparse Agent Mesh coordination without assuming scale guarantees.
Read articleAgent Mesh
Understand how Agent Mesh and Collective Runtime handle incomplete peer views, partition policies, reconciliation and restrictions on irreversible effects.
Read articleAgent Mesh
Learn how the AgentPlat Rooms Mesh bridge projects objectives and work, returns results and preserves the separation between Room policy and Mesh authority.
Read articleAgent Mesh
Understand what a collective quorum certificate establishes under an authenticated membership view, and why certification is separate from external-action authority.
Read articleAgent Mesh
Learn how Morphogenesis V4 exchanges signed outcome attestations without raw mission content, while receiving peers retain local compatibility and governance checks.
Read articleControl execution and understand recovery, approvals and audit records.
Human governance
Understand artifact review, task approval and execution checkpoints, and where to enforce human approval in an AI agent workflow.
Read articleReliable execution
Explore stable work identity, durable results and lease-based recovery for AI agent workflows, including the limits of retrying external effects.
Read articleGovernance & reliability
Understand the checks that belong before protected agent effects, including current assignment, policy, fencing, approvals and effect-specific idempotency.
Read articleGovernance & reliability
Learn where AgentPlat inference controls constrain model invocation, how they relate to provider checkpoints and why calls outside the integration path remain separate.
Read articleGovernance & reliability
Learn what to retain in an AgentPlat audit trail, how causal checkpoints support replay and why reconstructing decisions must not blindly repeat external effects.
Read articleGovernance & reliability
Design the identity boundary for AgentPlat tenants and actors, including resource authorization and tests for cross-tenant access and participant impersonation.
Read articleGovernance & reliability
Understand how AgentPlat Trust narrows local eligibility, where opt-in Mesh and inference adapters apply, and why a favorable profile is not action permission.
Read articleGovernance & reliability
Distinguish revision checks from idempotency in AgentPlat, and design retries that reject stale writes without turning the same logical operation into new work.
Read articleGovernance & reliability
Explore AgentPlat’s incident lifecycle, coordinated revocation and forensic preservation, including why recovery needs new credentials and an explicit authority transition.
Read articleGovernance & reliability
Use AgentPlat recovery receipts to connect failure detection, fencing, eligibility checks and resumption, and distinguish recovered progress from repeated activity.
Read articleGovernance & reliability
Plan a restore test that checks AgentPlat artifacts, operational checkpoints and current membership or lease authority before resumed workers perform effects.
Read articleEvolve a collective through governed proposals, decisions, execution boundaries and evidence.
Agent Morphogenesis
Understand AgentPlat Morphogenesis, including organizational proposals, scoped decisions, execution authorization and receipts from the subsystems that own changes.
Read articleAgent Morphogenesis
Choose policy, agent, human, collective or composite review routes in Morphogenesis while preserving scope, current mandates and proposer–reviewer independence.
Read articleAgent Morphogenesis
Explore AgentPlat’s governed operator catalog, persisted execution plans and compensation records when an organizational change crosses multiple subsystem boundaries.
Read articleAgent Morphogenesis
Understand Morphogenesis V3 strategy learning, multidimensional evidence, review and catalog transitions without turning a learned recommendation into execution authority.
Read articleAgent Morphogenesis
Explore Morphogenesis V5 from evidenced catalog gaps to declarative candidates, adversarial simulation, independent certification, review and experimental canaries.
Read articleAgent Morphogenesis
Understand the governed Agent Genesis lifecycle from an evidenced capability gap to a reviewed lineage, including sandbox, probation and separate membership admission.
Read articleAgent Morphogenesis
Understand Morphogenesis V7 successor organizational epochs, separate subsystem receipts and why compensation across multiple owners may remain partial.
Read articleAgent Morphogenesis
Recover effects from a Morphogenesis proposal that loses a concurrent successor race, with owner receipts, stable operation identities and explicit cleanup.
Read articleAgent Morphogenesis
Understand Morphogenesis V8 successor constitutional epochs, reviewed amendments, predecessor evidence and isolation of competing authority lineages.
Read articlePlan missions, preserve continuity and govern organizational change.
Collective systems
Understand how AgentPlat Collective Runtime connects mission intent, planning, teams, execution and recovery through explicit policy-bound lifecycle transitions.
Read articleCollective systems
Distinguish plans, eligible assignments and effect authority in AgentPlat, including how expiry and reassignment preserve the reason work changed owners.
Read articleCollective systems
Learn how AgentPlat continuity binds checkpoints to missions, assignments and membership, and what must be revalidated when a replacement peer resumes work.
Read articleCollective systems
Understand AgentPlat’s asynchronous governed-runtime transitions, durable stores and idempotency ledger for pause, recovery and safe-stop operations.
Read articleCollective systems
Understand versioned coordination strategies and governed team split, merge or federation, including topology lineage and activation evidence.
Read articleEvaluate the framework with observable tasks and honest evidence.
Adoption & evaluation
Plan an AgentPlat evaluation with recorded prerequisites, concrete collaboration tasks, honest failure reporting and separate evidence for usability and software behavior.
Read articleAdoption & evaluation
Plan an AgentPlat simulation campaign with bounded workers, evaluator-owned checkpoints, cross-shard delivery and attributable artifacts.
Read articleAdoption & evaluation
Register a multi-agent benchmark before execution and preserve source, topology, seeds, faults, checkpoints and raw results separately from the final interpretation.
Read articleAdoption & evaluation
Distinguish AgentPlat source implementation, package availability, deterministic checks, staging profiles and observed operational outcomes during technical evaluation.
Read articleAdoption & evaluation
Choose between an AgentPlat source checkout and registry packages, record exact coordinated versions and interpret next/latest tags without assuming current parity.
Read articlePublished by AgentPlat, the open-source TypeScript framework for persistent human-agent collaboration. These guides explain design decisions and documented examples, with links to primary sources. Illustrative scenarios are identified as examples; documented checks are not presented as production guarantees.
Implementation details live in the documentation and source repository. Each article includes publication and, when revised, update dates.