54 docs tagged with "tool"

Use when reading the current Player's Agent default overrides (LLM provider, settings, etc.) — the JSON shape used by the profile UI.

Use when you need to clear all Player Agent default overrides so the platform's admin defaults apply.

Use when you need to replace the current Player's Agent default overrides. Same JSON shape the profile UI uses. Affects new Agents and unset fields on existing ones.

Use when you need persisted Agent activity, optionally scoped to a runtime conversation. Overseer-only.

Use when you need to create or upsert a deterministic Agent runtime conversation. Overseer-only.

Use when you need to delete a specific Agent runtime conversation. Overseer-only.

Use when you need to duplicate an existing Agent's configuration (settings, mounted files, defaults). Overseer tool — operates on managed Agents.

Use when you need to stop a running Agent turn. Overseer-only.

Use when you need current conversation-scoped runtime stream state for a managed Agent. Overseer-only.

Use when you need persisted Agent messages, optionally scoped to a runtime conversation. Overseer-only.

Use when you need to check an Agent's runtime reachability, authentication, queue depth, and processing state. Diagnostics for Agent chat health.

Use when you need to send one text message to a managed Agent's chat runtime. Overseer-only.

Use when you need to start (or reconnect) an Agent's runtime chat session on the current Marquee. First step before sending messages.

Use when you need to upload an Artefact (file/report/plan/result) for the current Agent. Persists to Rails AND mirrors into FIBE_WORKSPACE_PATH if set.

Use when you need to switch session-scoped Fibe API key and/or domain mid-session for multi-tenant work. Validates new creds via /api/me ping by default.

Use when you need to invoke a Fibe MCP tool that is registered server-side but not advertised in the current client tool surface (typical when running --tools core).

Use when you need to verify Fibe MCP connectivity, API key validity, and current player profile. First call when troubleshooting auth or environment selection.

Use when you need to fetch one feedback entry by id, including the full Player comment, source reference, and context.

Use when you need to list all feedback entries (Player comments on Artefacts/Mutters/etc.) for the current Agent. Required reading after major milestones.

Use when you need to search GitHub repos across ALL the Player's connected GitHub App installations. Aggregates and deduplicates across orgs/accounts.

Use when you need to mint a short-lived GitHub installation access token for a specific repository. Auto-resolves the correct GitHub App installation.

Use when you need to create a new repo on the Player's connected Gitea (managed or external) and atomically register it as a Fibe Prop.

Use when you need to create a new GitHub repository under the Player's OAuth-connected GitHub account and register it as a Fibe Prop.

Use when bootstrapping a greenfield app in one call — creates one or more repos/Props from a template or GitHub repo snapshot, an app-owned template version, a deployed Playground, waits for running, and links it locally to /app/playground.

Use when you need to read the cobra CLI Long help for any fibe subcommand (flag descriptions, examples, payload shapes).

Use when deploying an existing Fibe-compatible Compose config from inline YAML, a local file, or a GitHub repository config file without creating new source repos first.

Use when you need to get filesystem-level info about one local Playground (name, path, playspec, mounts) without an API call.

Use when you need to symlink a local Playground's mounts into a working directory (default /app/playground) so the Agent has direct file access. Brownfield handshake.

Use when you need to create or update one Memory grounded in a local source conversation. Persistent agent memory tied to evidence (groundings + conversation snapshot).

Use when you need to stream Agent events live as MCP progress notifications. Bounded by duration and max_events. For Overseer waiting on Agent output.

Use when you need to list Agent-produced events (messages, activities, mutters, artefacts) with standard pagination. Snapshot mode — for one-shot reads.

Use when you need to post one short progress/proof/blocker/problem/milestone note as the current Agent. The dedicated Agent progress channel.

Use when you need to retrieve one Agent's mutter stream — the "thinking out loud" feed of progress, proof, problem, and blocker notes — with optional query/status/severity/playground filters.

Chain multiple Fibe tool calls into one round-trip with JSONPath bindings. Use when latency matters or multi-step jobs (create→wait→link) need atomicity.

Use when you need to re-fetch a cached fibe_pipeline result by id, optionally projecting a specific JSONPath, within 5 minutes of the original run.

Use when you need to run a single Playground lifecycle action — rollout, hard_restart, stop, start, retry_compose, enable_maintenance, or disable_maintenance. Destructive; requires confirm:true.

Use when you need to fetch comprehensive Playground diagnostics — service/container names, ports, paths, labels, status, urls, recent logs — when troubleshooting a deployment.

Use when you need to fetch a bounded log snapshot for a single Playground service. Async on Rails — returns terminal payload after polling.

Use when you need to stream a Playground service's live logs as MCP progress notifications until duration elapses or max_lines reached. For waiting on a specific log pattern.

Use when an existing deployed playground needs to be transformed end-to-end to a different stack, service shape, source layout, or prop set while preserving its id. Single-call brownfield analog of fibe_greenfield_create. Authors a new template inline if needed and provisions Gitea-backed private Props on the fly for new repos.

Use when you need to block-and-poll until a Playground reaches a target status (running, stopped, has_changes, destroyed, etc.). Required after rollout/start/restart/job triggers.

Use when you need to verify Fibe's view of multiple GitHub repository URLs in one call — accessibility, App installation status, default branch, etc.

Use when you need to permanently delete a Fibe resource by id (or name for playgrounds/tricks/playspecs/props/marquees/agents). Destructive — requires confirm:true.

Use when you need to fetch a single Fibe resource by id, or by name for playgrounds, tricks, playspecs, props, marquees, and agents. Includes artefact_attachment download.

Use when you need to list any flat Fibe resource (playground, prop, marquee, agent, secret, webhook, audit_log, memory, artefact, etc.) with schema-validated filters.

Use when you need a single tool for create/update and resource-scoped operations (attach, mirror, sync, fork, trigger, rerun, action, etc.) across Fibe resources. Schema-validates locally before API call.

Use when no native MCP tool fits and you need a last-resort escape hatch to invoke an arbitrary fibe CLI command when no MCP tool fits. Avoid; native MCP tools and fibe_call are safer.

Use before calling Fibe mutation/list tools that need schemas. Authoritative source for resource/operation JSON schemas. Required reading before calling fibe_resource_mutate, fibe_resource_list, or any tool with a payload field.

Use when you need to dashboard the current player's resource counts, quota usage, rate-limit budget, and subscription plan in one call.

Use when an advanced template-author or admin workflow needs the hidden lower-level primitive for patching, overwriting, or switching template versions, including Trick reruns and global template promotion.

Use when you need to bootstrap and launch a Playground directly from an existing import template, without creating a new repo. Greenfield-lite.

Use when you need to search the Fibe Import Template catalog (own + public + team) by full-text or PostgreSQL regex before greenfield/launch.

Use when you need to discover the full Fibe MCP tool surface (advertised + hidden). Required first step before fibe_call when targeting a tool not visible in ToolSearch.

Use when you need to update the current Agent's display name. Trigger when the conversation topic changes materially.