docs(#336): tool results bounded by maxResultSize — Tools (Result size), Configuration (defaults.tools), Observability, Changelog — PR 2
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#268): generators and naming — CLI (full names, generate workflow, create), Tools (names, timeouts), Naming (new), Components, App, Pauses, Testing, Knowledge bases, Changelog — PR 4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#335): lease fencing at expiry, unique tool-call ids — Several-instances, Tools, Changelog
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#267): interpreter outcomes, a decision-model interpreter, child pause records — Pauses, Testing, Tools, Workflow-as-a-step, Changelog — PR 4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#267): reply interpreters, pause answer records and retention — Pauses, Tools, Observability, Several-instances, Testing, Changelog
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#262): concurrency and unit runners — Testing (TestConcurrency, unit runners), Test-split (new), Tools, Components, CLI, Changelog — PR 3
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#267): pauses and resume — Pauses (pending, resume by id, who answers, a pause's life, concurrent answers), Pause seam (S35), Tools, Testing, App, Several instances, Observability, Changelog — PR 2
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#267): channel roles and discovery — Tools (approval channels: @ChannelRole, implementations in @Application providers, send and ChannelOutcome, assembly codes), Pause-seam (no channelClasses, send once, decided), Components, Agents, Testing, Changelog — PR 1
docs(#261): budgets and limits at run time — FinOps (budgets checked, the ledger, steps and visits, eval), Configuration, App, Changelog and the pages that named settings().limits — PR 5
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#280): SpendRecorder (S43), ctx.pause removed until #267 (S44), a tool call's signal (S45)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#280): the base ExecutionContext — Context and Authorization, Components, Tools, Judges, Testing, FinOps, Saga, Changelog
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#272): the flow as paths — Workflow, Path scope, Rollback, ADR-0019, routing, components, agents, CLI, changelog
Every page in the path grammar (#272, #270, #269): Workflow, Advanced-Routing (routing patterns),
Routers, Components, Example, Standard-DTOs (Dto suffix, wrappers), Agents (input message, answer
tools), CLI (agboss check writes types), Model-providers (ProviderSort), new Path-scope, Rollback,
ADR-0019 and Changelog pages; ADR-0014/0017 superseded; the other pages' snippets on v2.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XQGj6TfWHuvSMazca2Fw4p
docs(#249): rename to AGBoss (ADR-0018)
docs(#240): rewrite the wiki against the current API; every ts block compiles
Pages using the old API (route().to(), instructions:, graphcompose/core,
@graphcompose/core, app.run, routeOne, fork(), Suite/Dataset, BaseJudge, ...)
are rewritten from the current code; every ts block now compiles with
`node scripts/check-docs.mjs --wiki <checkout>` and every link between pages
names an existing page. Removes the leftover patch scripts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NSCfutKyDuav7Yy364CayR
docs: fix testing import path
docs: add testing examples and limitations to core entities
docs(#150): the agent's own loop — Agents page, ADR-0015, pause seam one call per pause, reasoning attempts removed
docs(#141): @WorkflowStart / @WorkflowFinish, standard DTOs by the hierarchy rule, RagRetrieval { results }, ToolCallApprovalDecision, PlainText
docs(#116): job-scout on the flow graph — Workflow, Routers, ADR-0014; hub, routers and budget removed across pages
docs(#118): DTOs — Components, Tools, Standard DTOs, ADR-0013
docs(#109): component rules, MCP
docs(#107): file conventions
docs(#91): graphInject package, Example, ADR-0011
docs(#87): components, ADR-0009
docs(#85): who can use which tool
docs(#20): agents with tools
docs(#18): Tools page, ADR-0003