Agent manifest (`kind: Agent`)¶

Package: mas-runtime · Schema: agent.schema.yaml · apiVersion: mas/v1

An agent manifest (agent.yaml) declares one LLM actor: tools, skills, design pattern, governance/control, and observability settings. A MAS manifest references one or more agents; overlays patch agents without duplicating the base file.

Terms: glossary.md · Hub: README.md.

Top-level fields¶

Every manifest starts with four required top-level fields.

Field	Type	Required	Description
`apiVersion`	`string`	yes	Must be `mas/v1`.
`kind`	`string`	yes	Must be `Agent`.
`metadata`	object	yes	Identity and tagging. See `metadata` fields.
`spec`	object	yes	Runtime behaviour. See `spec` fields.

Extension properties (x-*) are allowed at the top level and are ignored by the runtime (they carry UI metadata, canvas state, etc.).

Minimal example:

apiVersion: mas/v1
kind: Agent
metadata:
  name: broker
spec:
  description: "Broker agent. Routes requests to specialists."

`metadata` fields¶

Field	Type	Required	Default	Description
`name`	`string`	yes	—	Agent ID — must match the workflow node id that references this agent.
`description`	`string`	no	`""`	Human-readable description of this agent.
`version`	`string`	no	`"0.1.0"`	Semver string (`major.minor.patch`).
`tags`	`string[]`	no	`[]`	Free-form tags for filtering and grouping.

Extension properties (x-*) are allowed and ignored by the runtime.

`spec` fields¶

spec.description is the only required field.

Field	Type	Required	Default	Description
`description`	`string`	yes	—	Routing-facing one-liner for delegation tools, AgentCard, and registry discovery. Written from the delegator's perspective. Not injected into the LLM system prompt — use `context.*` for prompt text.
`context`	`object`	no	—	Named chunks injected into the system prompt. Keys are author-defined (`role`, `intent`, …). See `ContextChunk`.
`params`	`object`	no	`{}`	Free-form string key/value params forwarded to infra middleware — not consumed by the kernel.
`models`	`ModelEntry[]`	no	`[]`	Preferred models for this agent. See `ModelEntry`.
`design_pattern`	object \| null	no	`react`	Reasoning strategy. See `DesignPattern`.
`context_manager`	object	no	`summarising`	Context window management. See `ContextManager`.
`assembler`	string \| object \| null	no	`assembler`	Builds the LLM `messages[]` from context.
`memory_seed`	`MemorySeedEntry[]`	no	—	Documents to pre-index into memory at startup. See `MemorySeedEntry`.
`memory`	string \| string[] \| object	no	—	Memory backend configuration. See `MemoryConfig`.
`working_memory`	object	no	—	Cross-turn history persistence for delegated agents. See `WorkingMemory`.
`checkpoint`	object	no	`{mode: none}`	Session checkpoint cadence, retention, portability and automatic resume. See checkpoint.md.
`skills`	`string[]`	no	`[]`	Skills to activate. Name-only or `@library/name`. Resolution: app-local → libraries → packages.
`tools_ref`	`string` \| null	no	`null`	Logical tool-set name resolved by the infra `ToolRegistry` (e.g. `web-tools`). No paths or extensions.
`tools`	`Tool[]`	no	`[]`	Per-agent tool declarations — three forms. Additive with `tools_ref`. See `Tool`.
`behavior`	object	no	—	Runtime capability flags. See `Behavior`.
`governance`	`GovernanceBinding`	no	`{}`	Governance plugin list. See `GovernanceBinding`.
`hitl_contract`	`string`	no	`registry`	Human-approval Protocol plugin. An explicit `InstantiationOptions.hitl_contract` instance overrides this binding, preserving interactive CLI adapters.
`user_io_contract`	`string`	no	`registry`	User progress-update Protocol plugin. Overridden the same way as `hitl_contract`.
`llm`	`LlmBinding`	no	`{}`	Engine overrides (model, temperature, …). See `LlmBinding`.
`budget`	object	no	`{}`	Call-count ceilings (`max_llm_calls`, `max_tool_calls`) enforced by the engine. Omitted = no limit. See `Budget`.
`control`	`ControlBinding`	no	`{}`	Control-plane plugin configs. See `ControlBinding`.
`observability`	`ObservabilityBinding`	no	`null`	Observability sink plugin list. See `ObservabilityBinding`.
`context_sources`	`ContextSourcesBinding`	no	`null`	Context source plugins, including the skill-engine backend.

Sub-schemas¶

Assembler¶

Used by: spec.assembler

The schema accepts a plugin-name string, null, or an object with type or ref. Its optional params object is plugin-specific and currently defines token_budget, always_reassemble, and emit_segments.

ContextSourcesBinding¶

Used by: spec.context_sources

An ordered list of context-source plugin ids (native, adk, or langchain). An entry may be a bare id or a single-key object with base_dir, auto_inject, pin_activated, allow_unload, and auto_load options. auto_inject and allow_unload default to false. allow_unload is off because activate_skill(name, unload=true) is not in the Agent Skills spec.

ContextChunk¶

Used by: spec.context.<key>

Each value under spec.context is a named chunk injected into the system prompt under [key]. Two forms:

Form	Type	Description
Inline string	`string`	Injected as literal text, unless the value resolves to an existing file path (e.g. `prompts/role.md` or `./prompts/role.md`).
File reference	`{ ref: string }`	Always loads the file at `ref`. Raises `ContextRefNotFoundError` at bootstrap when the path is missing.

Example:

context:
  role: |
    You are a telemetry analyst…
  intent:
    ref: "./prompts/intent.md"

ModelEntry¶

Used by: spec.models[]

Field	Type	Required	Default	Description
`model`	`string`	yes	—	LiteLLM-style model string, e.g. `vertex_ai/gemini-3-pro-preview`.
`id`	`string`	no	`"main"`	Logical model ID within this agent. Use `main` for the primary model.
`temperature`	`number` [0.0–2.0]	no	`0.7`	Sampling temperature.
`max_tokens`	`integer` ≥ 1	no	unset	Output-token budget; omitted means no limit is sent.
`max_completion_tokens`	`integer` ≥ 1	no	unset	Explicit answer + reasoning cap; the request never carries both token-limit fields.
`on_truncation`	string or object	no	`warn`	Action when the provider returns `finish_reason: length`; see Output-token limits.
`context_window`	`integer` ≥ 1	no	—	Input context-window size used for model-aware assembly.

Do not put api_base or api_key_env here — those belong in the flavour/infra manifest.

Example:

models:
  - model: vertex_ai/gemini-3-pro-preview
    temperature: 0.3
    max_tokens: 4096

Output-token limits¶

By default, requests send neither max_tokens nor max_completion_tokens; the provider's server default applies. A configured budget is lowered per call so the estimated prompt plus completion fits context_window. The effective cap is also bounded by the lower of infra spec.models.generation.max_output_tokens and the model catalog's max_output_tokens.

Setting	Meaning
`max_tokens`	Output budget; reasoning models may map it to `max_completion_tokens`.
`max_completion_tokens`	Explicit answer + reasoning budget; takes precedence over `max_tokens`.
`on_truncation`	`ignore`, `warn` (default), `error`, or `escalate`. Object form accepts `factor` (default 2), `max_tokens`, and `retries` (default 1). Retries count against `spec.budget.max_llm_calls`.

The CLI alias expands to agent:spec.models[*].max_tokens=N; for example, mas-ctl chat agent.yaml --max-tokens 4096 is equivalent to --override 'agent:spec.models[*].max_tokens=4096'. Explicit --override arguments are applied afterward. Environment variables MAS_LLM_MAX_TOKENS, MAS_LLM_MAX_COMPLETION_TOKENS, and MAS_LLM_ON_TRUNCATION fill in settings not declared on the model row.

LLM traces include finish_reason, the final max_tokens budget, completion token usage, and truncation_retries. See also infra.md — LLM output-token policy and CLI overrides.

DesignPattern¶

Used by: spec.design_pattern

Selects the agent's reasoning strategy (defaults to react when absent).

Field	Type	Required	Description
`type`	`string`	mutually exclusive with `ref`	Builtin shorthand: `react` \| `cot` \| `cot_native` \| `plan_execute` \| `tree_of_thoughts` \| `accumulator` and role aliases.
`ref`	`string`	mutually exclusive with `type`	Unified plugin locator — bare name, `./local/path`, `module://pkg.Class`, or `oci://registry/img:tag`.
`params`	`object`	no	Plugin-specific parameters.
`config`	`object`	no	Alias for `params` (overlay compatibility).

Example:

design_pattern:
  type: react

# or a custom plugin:
design_pattern:
  ref: module://my_pkg.patterns.MyCoT
  params:
    max_steps: 10

ContextManager¶

Used by: spec.context_manager

Controls context window management strategy (defaults to summarising).

Field	Type	Required	Description
`type`	`string`	mutually exclusive with `ref`	Builtin: `stack` \| `sliding-window` \| `summarising` \| `full-history`.
`ref`	`string`	mutually exclusive with `type`	Plugin locator — same forms as `design_pattern.ref`.
`params`	`ContextManagerParams`	no	Strategy-specific parameters. See `ContextManagerParams`.
`memory`	`string[]`	no	Memory types whose content is auto-injected into the system prompt as context facets, e.g. `[semantic]`.

Example:

context_manager:
  type: sliding-window
  params:
    window_size: 20

ContextManagerParams¶

Used by: spec.context_manager.params

Constructor kwargs forwarded to the ContextManagerPlugin. All fields are optional.

Field	Type	Description
`max_turns`	`integer` ≥ 1	Sliding-window — maximum prior user turns; alias for `keep_turns`.
`window_size`	`integer` ≥ 1	Sliding-window — alias for `keep_turns`.
`max_messages`	`integer` ≥ 1	Stack context manager — cap on total past messages.
`summary_threshold`	`integer` ≥ 0	Summarising — token estimate that triggers compaction; `0` uses the derived model context window.
`keep_turns`	`integer` ≥ 1, default `10`	Summarising — recent user turns kept verbatim.
`hysteresis_ratio`	number [0–1], default `0.1`	Summarising — reuse the cached summary until near the safety-bounded hard budget.
`summarizer`	string \| object	Sub-plugin: `llm` or `drop`; object form can set model and instructions.
`working_memory_messages`	`integer` ≥ 1, default `20`	Assembly cap on in-turn working-memory messages.
`max_in_turn_messages`	`integer` ≥ 1	Alias for `working_memory_messages`.
`trimmer`	object	Assembly-time token trimming; accepts `max_tokens`/`token_budget`, `reserve_tokens`, and `safety_margin_ratio`.
`working_memory_ref`	`string`	no

Nested trimmer and summarizer keys are defined in the context-manager parameter schemas.

MemorySeedEntry¶

Used by: spec.memory_seed[]

Pre-indexes a document into the memory backend at startup. Useful for demos and tests that need pre-populated memory without a persistent store.

Field	Type	Required	Description
`content`	`string`	yes	Document text to index.
`key`	`string`	no	Lookup key.
`source`	`string`	no	Source label / filename.
`id`	`string`	no	Explicit document ID.
`text`	`string`	no	Alias for `content`.

Example:

memory_seed:
  - source: "runbook.md"
    content: |
      ## Restart procedure
      1. Drain traffic…

MemoryConfig¶

Used by: spec.memory

Three forms are accepted:

Form	Example	Description
String shorthand	`memory: semantic`	Resolves to the corresponding plugin bundle.
Array shorthand	`memory: [semantic, episodic]`	Activates multiple named bundles.
Object form	`memory: {enabled: true, types: [...]}`	Full configuration — described below.

Object form fields:

Field	Type	Default	Description
`enabled`	`boolean`	`true`	Master switch — `false` loads no memory plugins.
`types`	`MemoryType[]`	`[]`	Ordered list of memory layers. See `MemoryType`.
`persistence`	object	—	Session transcript persistence. See `MemoryPersistence`.
`search`	object	—	Semantic retrieval configuration. See `MemorySearch`.
`citations`	object	—	Citation formatting for search results. See `MemoryCitations`.
`sync`	object	—	File-watching and re-indexing policy. See `MemorySync`.
`overflow_retry`	object	—	Compaction retry on context overflow. See `MemoryOverflowRetry`.
`qmd`	object	—	QMD (Qualitative Memory Database) backend. See `MemoryQMD`.

MemoryType¶

Used by: spec.memory.types[]

Field	Type	Required	Default	Description
`name`	`string`	yes	—	Layer name: `session` \| `episodic` \| `semantic` \| `working` \| `procedural`.
`backend`	`string`	no	`"in-memory"`	Backend: `in-memory` \| `file` \| `sqlite-vec` \| `redis` \| `remote_tool`.
`params`	`MemoryBackendParams`	no	`{}`	Backend constructor kwargs. See `MemoryBackendParams`.

MemoryBackendParams¶

Used by: spec.memory.types[].params

Field	Type	Description
`path`	`string`	Filesystem path (file backend).
`url`	`string`	Connection URL (redis, remote_tool).
`collection`	`string`	Collection / table name.
`host`	`string`	Hostname.
`port`	`integer`	Port number.
`database`	`string`	Database name.

MemoryPersistence¶

Used by: spec.memory.persistence

Field	Type	Default	Description
`backend`	`string`	`"none"`	`none` \| `file` \| `sqlite` \| `redis`.
`path`	`string`	`""`	Base directory. Supports `{agent_id}` and `{session_id}` placeholders. Default: `$XDG_DATA_HOME/mas/agents/{agent_id}/sessions/`.
`auto_save`	`boolean`	`true`	Persist after each turn.
`auto_load`	`boolean`	`true`	Load session on bootstrap.

MemorySearch¶

Used by: spec.memory.search

Field	Type	Default	Description
`enabled`	`boolean`	`false`	Enable semantic search injection into context.
`max_results`	`integer`	`6`	Maximum retrieved chunks per query.
`min_score`	`number`	`0.35`	Minimum similarity score threshold.
`chunking.tokens`	`integer`	`400`	Chunk size in tokens for indexing.
`chunking.overlap`	`integer`	`80`	Overlap tokens between chunks.
`hybrid.enabled`	`boolean`	`true`	Enable hybrid search (vector + full-text).
`hybrid.vector_weight`	`number`	`0.7`	Vector score weight in hybrid ranking.
`hybrid.text_weight`	`number`	`0.3`	Full-text score weight in hybrid ranking.
`hybrid.mmr_enabled`	`boolean`	`false`	Max Marginal Relevance re-ranking.
`hybrid.mmr_lambda`	`number`	`0.7`	MMR lambda (diversity vs relevance).
`hybrid.temporal_decay_enabled`	`boolean`	`false`	Apply time-decay to scores.
`hybrid.temporal_decay_half_life_days`	`integer`	`30`	Half-life for temporal decay.
`cache.enabled`	`boolean`	`true`	Cache search results.
`cache.max_entries`	`integer`	`128`	Maximum cached search results.

MemoryCitations¶

Used by: spec.memory.citations

Field	Type	Default	Description
`mode`	`"auto"` \| `"on"` \| `"off"`	`"auto"`	Citation mode: `auto` (include when useful), `on` (always), `off` (strip source info).
`max_snippet_chars`	`integer`	`700`	Maximum characters per search result snippet.

MemorySync¶

Used by: spec.memory.sync

Controls when workspace memory files (MEMORY.md, memory/*.md) are re-indexed.

Field	Type	Default	Description
`on_session_start`	`boolean`	`true`	Re-index when a session begins.
`on_search`	`boolean`	`false`	Re-index before each search.
`interval_seconds`	`integer`	`300`	Periodic re-index interval in seconds (0 = disabled).
`delta_messages`	`integer`	`50`	Re-index after this many new messages.
`delta_bytes`	`integer`	`100000`	Re-index after this many bytes of new content.
`post_compaction_force`	`boolean`	`true`	Force re-index after compaction.
`file_watch.enabled`	`boolean`	`true`	Enable file change detection.
`file_watch.patterns`	`string[]`	`["MEMORY.md","memory/*.md"]`	Glob patterns to watch.
`file_watch.debounce_ms`	`integer`	`1500`	Debounce interval for file change events.

MemoryOverflowRetry¶

Used by: spec.memory.overflow_retry

When an LLM call fails because the context exceeds the window, automatically compact and retry.

Field	Type	Default	Description
`enabled`	`boolean`	`false`	Enable overflow detection and retry.
`max_retries`	`integer`	`2`	Maximum compaction + retry attempts.
`aggregate_timeout_seconds`	`number`	`60.0`	Maximum total time across all retries.
`budget_reduction_factor`	`number`	`0.7`	Multiply token budget by this factor on each retry.

MemoryQMD¶

Used by: spec.memory.qmd

Alternative search via external qmd binary (Qualitative Memory Database).

Field	Type	Default	Description
`enabled`	`boolean`	`false`	Enable QMD as a retriever backend.
`binary_path`	`string`	`""`	Path to `qmd` binary (auto-detected if empty).
`index_path`	`string`	`""`	Path to QMD index directory.
`search_mode`	`string`	`"search"`	QMD search mode.
`max_snippet_chars`	`integer`	`700`	Maximum characters per snippet.
`timeout_seconds`	`number`	`4.0`	Search timeout in seconds.

WorkingMemory¶

Used by: spec.working_memory

Controls whether a delegated agent's committed conversation history (user/assistant turns) survives across separate delegation calls within the same mas-ctl session, keyed by (session_id, agent_id).

Distinct from spec.memory: that's the MemoryContract retrieval-store subsystem (semantic/episodic/procedural search); this is the raw turn buffer a delegated agent falls back on.

Field	Type	Default	Description
`persistent`	`boolean`	`true`	`true`: reuse committed turn history across delegate calls within the same session. `false`: clear working memory before every delegate call — for agents that must be stateless per call (e.g. a formatter or translator).
`compaction`	`WorkingMemoryCompaction`	—	How much history to keep as it grows. See `WorkingMemoryCompaction`. Facade over `spec.context_manager` — set `context_manager` directly instead for lower-level control (takes precedence if both are set).

Context ID: the delegating agent may pass context_id alongside task on a delegate_to_<agent_id> call to select an independent working-memory bucket for that peer (e.g. "trip-paris" vs "trip-tokyo" within one session). Omit it to use the session's default bucket.

WorkingMemoryCompaction¶

Used by: spec.working_memory.compaction

Field	Type	Default	Description
`strategy`	`"keep_recent"` \| `"sliding_window"` \| `"summarize"`	`"keep_recent"`	Compaction strategy (see below).
`max_messages`	`integer` ≥ 1	—	`keep_recent` — cap on total committed messages.
`window_size`	`integer` ≥ 1	—	`sliding_window` — number of recent exchange pairs to keep.
`summary_threshold`	`integer` ≥ 1	`4000`	`summarize` — estimated-token threshold before compaction triggers.
`keep_turns`	`integer` ≥ 1	`10`	`summarize` — recent exchange pairs kept verbatim alongside the summary.

Strategies:

keep_recent — cap total messages at max_messages, no LLM call.
sliding_window — keep the last window_size exchange pairs, no LLM call.
summarize — compress older turns into one summary block via this agent's own model, keeping keep_turns recent pairs verbatim. Degrades to keep_recent when no live model is available.

Example:

working_memory:
  persistent: true
  compaction:
    strategy: summarize
    summary_threshold: 4000
    keep_turns: 10

Tool¶

Used by: spec.tools[]

Three forms are accepted. All are additive with tools_ref.

Form A — manifest reference (recommended)¶

References a kind: Tool manifest file or a ToolBundle entry.

Field	Type	Required	Default	Description
`ref`	`string`	yes	—	Path to a `kind: Tool` manifest (`./tools/calc.tool.yaml`) or a ToolBundle entry (`bundle://web-tools/fetch-page`).
`priority`	`integer`	no	`100`	Registration priority (higher = loaded first).
`params`	`object`	no	`{}`	Init kwargs merged into the referenced tool's `spec.impl.params`.

Form B — inline anonymous¶

Inline declaration for a Python class, remote tool, or OpenAPI endpoint.

Field	Type	Required	Default	Description
`module_path`	`string`	yes	—	Dotted Python module path or relative file path (`./tools/my_tool.py`).
`kind`	`"python"` \| `"remote_tool"` \| `"openapi"`	no	`"python"`	Implementation kind.
`class_name`	`string`	no	—	Class name in the module. Auto-discovered when omitted.
`priority`	`integer`	no	`100`	Registration priority.
`params`	`object`	no	`{}`	Optional tool-specific init params.

Form C — semantic name¶

A bare string resolved by the active infra/local tool catalog (e.g. web-search, calculator, memory-search).

Example:

tools:
  - ref: ./tools/calculator.tool.yaml          # Form A
  - module_path: library-samples/tools/calc.py  # Form B
    class_name: CalcTool
  - web-search                                  # Form C

Behavior¶

Used by: spec.behavior

Structural/semantic flags that affect which system tools and capabilities are exposed to the agent's LLM at runtime.

Field	Type	Default	Description
`share_reasoning`	`boolean`	`false`	When `true`, the `send_to_caller` tool exposes an optional `reasoning_context` parameter. The sub-agent may provide a concise summary of HOW it reached its answer (key evidence, decision points, confidence signals), forwarded to the orchestrator as `result["reasoning"]`. Enable only for trusted internal agents where context leakage is acceptable.
`delegation_style`	`"typed"`	`"typed"`	Controls which delegation tools are registered. `typed`: `delegate_to_<id>` tools derived from the MAS workflow.

GovernanceBinding¶

Used by: spec.governance

Ordered plugin chain (BLOCK stops; ALLOW continues). Full page: governance.md. Retry / circuit / error_policy defaults: reliability.md.

Each entry is a bare plugin id or a single-key config object. error_recovery_plugin is YAML sugar that installs that plugin on the ingress chain (one classifier). error_policy is that plugin's config — the kernel only applies the returned GovernanceAction.

governance:
  - gov_no_undeclared_tool
  - retry_on_error:
      error_recovery_plugin: retry_on_error
      error_policy:
        transient: retry
        fatal: block

LlmBinding¶

Used by: spec.llm

EngineContract / LiveLlmEngine overrides. These complement spec.models[] — prefer models for per-agent routing and use llm for env-level or overlay-level adjustments.

Field	Type	Description
`model`	`string`	LiteLLM-style model string override.
`provider`	`string`	Engine provider hint (`mock`, `openai`, `azure`, …).
`temperature`	`number` [0–2]	Sampling temperature override.
`max_tokens`	`integer` ≥ 1	Maximum output tokens override.

Example:

llm:
  provider: mock

Budget¶

Used by: spec.budget

Call-count ceilings enforced by the runtime engine (LiveLlmEngine / BudgetTracker) during a run. Omit a field, or the whole object, for no limit — this is the default. Typically left unset on the agent manifest and set via an overlay instead (see with-hardened) rather than hardcoded per-agent. Distinct from spec.control.budget (token/cost accounting — see ControlBinding — which is not yet read by the engine).

Field	Type	Required	Default	Description
`max_llm_calls`	`integer` ≥ 0	no	unset	Maximum LLM calls allowed in a run.
`max_tool_calls`	`integer` ≥ 0	no	unset	Maximum tool calls allowed in a run.

Example:

budget:
  max_llm_calls: 50
  max_tool_calls: 100

ControlBinding¶

Used by: spec.control

Control-plane knobs. retry is applied by the runtime reliability layer (LLM HTTP + tool dispatch). circuit_breaker loads a library plugin (default threshold). budget and rate_limiter remain plugin configs. Set a key to null to disable. Full tables: reliability.md.

Key	Fields	Description
`budget`	`max_tokens: integer`, `max_cost_usd: number`	Token or cost budget enforcement.
`retry`	`llm` / `tools`: `max_attempts`, `backoff_s`, `backoff_multiplier`, `jitter`, `retry_on`, `require_idempotent`, `max_backoff_s`	Infra re-issue budget. LLM default 4 attempts on `transient`+`unavailable` (429/5xx, timeout, connect-refused); tools 2 attempts on `transient`+`unavailable` when idempotent.
`circuit_breaker`	`plugin`, `enabled`, `failure_threshold`, `reset_timeout_s`, `on`	Off unless this key is present. Loads the `threshold` circuit-breaker plugin.
`rate_limiter`	`requests_per_minute: integer`	Limit LLM call rate.

Example:

control:
  budget:
    max_tokens: 50000
  retry:
    llm:
      max_attempts: 4
      retry_on: [transient, unavailable]
  circuit_breaker:
    failure_threshold: 5
    reset_timeout_s: 30
    "on": [unavailable]
  rate_limiter:
    requests_per_minute: 30

error_policy and error_recovery_plugin live on a governance stanza, not under control — see reliability.md.

ObservabilityBinding¶

Used by: spec.observability

An ordered list of observability sink plugin ids. OSS sinks: native, otel.

Each entry is either a bare plugin id or a single-key config object:

observability:
  - native                                 # bare id — defaults
  - native:                               # with config
      path: ./traces
      events_file: events.jsonl
  - otel:
      output_path: ./otel-traces
      otel_file: spans.json

Unknown plugin ids fail at load time. Extended sinks (observe_sdk, etc.) are internal-only.

Schema source¶

# From installed package
python -c "from mas.lab.schemas.paths import runtime_schema_dir; print(runtime_schema_dir() / 'agent.schema.yaml')"

# From Web UI / controller (default port 8090)
curl http://localhost:8090/api/schemas/agent

Agent manifest (kind: Agent)¶

Top-level fields¶

metadata fields¶

spec fields¶

Sub-schemas¶

Assembler¶

ContextSourcesBinding¶

ContextChunk¶

ModelEntry¶

Output-token limits¶

DesignPattern¶

ContextManager¶

ContextManagerParams¶

MemorySeedEntry¶

MemoryConfig¶

MemoryType¶

MemoryBackendParams¶

MemoryPersistence¶

MemorySearch¶

MemoryCitations¶

MemorySync¶

MemoryOverflowRetry¶

MemoryQMD¶

WorkingMemory¶

WorkingMemoryCompaction¶

Tool¶

Form A — manifest reference (recommended)¶

Form B — inline anonymous¶

Form C — semantic name¶

Behavior¶

GovernanceBinding¶

LlmBinding¶

Budget¶

ControlBinding¶

ObservabilityBinding¶

Schema source¶

See also¶

Agent manifest (`kind: Agent`)¶

`metadata` fields¶

`spec` fields¶