Meaning System

:: How to use MeaningSystem — two modalities, one evidence layer

Your agent keeps agency. MeaningSystem provides proofs.

Maenen provides the trust, alignment and evidence layer for agent workflows.

MeaningSystem does not control — or need to control — your agent.
Your agent continues using its own tools, models, files and APIs as you see fit.
Purely deterministic, the system acts at checkpoints: shielding external evidence, checking alignment, detecting drift, evaluating legitimacy, and producing verifiable receipt evidence of the work done.

Open the Ledger → Browse the APIs →
01 · THE OPERATOR QUESTION

Explain how my agent uses the MeaningSystem

Initiation
First, your agent learns about the Maenen tools via the agent.md guide and holds a valid API key.

Your access key unlocks the whole available Maenen API toolsuite and MeaningSystem.

Basic relationship
You have your external agent — a coding agent, research agent, marketing agent, workflow agent, email/calendar agent, an OpenRouter / Claude / Gemini / GPT-based assistant, Hermes, OpenClaw, an n8n Zapier-style automation, or a company internal AI tool.

That same agent can call the MeaningSystem and individual Maenen APIs to gain efficiency insights, cost visibility and proof-backed checks.

02 · TWO MODALITIES

Semantic router, or singular instruments.

MODE A

MeaningSystem-led

Your agent submits its full prompt task to: POST /v1/run

Best for when you want to auto-select a workflow: compare two texts, summarise a brief, run an alignment check, create a meaning capsule, verify a receipt.

Good for: controlled presets · public demos · repeatable workflows · capsule generation · auditable product flows.

MODE B

Agent-led tool use

Your agent calls a known MeaningSystem route, or any single, multiple, or chained succession of Maenen APIs during a workflow. Best when your agent is already doing something larger — research, procurement, coding, investment, legal/compliance.

Good for: third-party agents · browser agents · research agents · enterprise workflows.

→   Developer 'builder' mode

03 · HOW AGENTS USE THE MEANINGSYSTEM MODES

Agents use MeaningSystem in two distinct modes:

MODE A · MeaningSystem-led

The agent sends your task or payload to the MeaningSystem for the requested processing. Raw submitted content is not intentionally placed in ordinary application logs; contract-required account, usage, receipt and evidence records may persist. The system then selects the correct route or preset:

  • secure_meaning_run_v1,
  • align_brief_v1,
  • compare_meaning_v1
etc.
As some agents will not know which checks to run, the MeaningSystem becomes a semantic router. So instead of the agent saying “call compress, then retrieval, then check drift, then integrity, then read the receipt”, it is essentially saying: “Here is the user goal and my output. Evaluate it.”.

→  The MeaningSystem orchestration is deterministic, not another LLM interpretation.

Submit the task. Receive the evidence.

First, your agent must familiarise itself with the MeaningSystem by reading the agent.md guide

Then the agent asks MeaningSystem to detect a route by entering its prompt task: “Here is my task — what MeaningSystem route should I use?”

Agent calls POST /v1/run
↓
Orchestrator selects approved route
↓
MeaningSystem executes route steps
↓
Receipts + timeline + DAG + capsule export
↓
Agent / User reviews performance

— Submit your payload, along these lines:

{ "preset_id": "secure_meaning_run_v1", "input": "User prompt or task payload here", "mode": "suggest_or_run", "client_context": { "agent_id": "agent.example.alpha", "session_id": "optional-client-session", "source": "user_agent" } }

— MeaningSystem automatically handles:

route selection · step execution · meaning/integrity checks · drift checks · receipt generation · cost accounting · proof surfaces · capsule export availability.


· THE DEVELOPER MODE B

MODE B · Agent-led

Your external agent selects known MeaningSystem routes, or individual Maenen tools as needed:

i.e.
An agent is prompted “research this topic X and summarize the strongest argument”.

The agent does its research externally, then might call MeaningSystem preset configurations, and/or also individual Maenen APIs during its task or post;

  • compress.v1 — to condense sources into an efficient storage format;
  • relevance.check — “does my summary answer the actual request?”;
  • faithfulness.check — “did I distort the source?”;
  • integrity.analyze — “does the final answer maintain the original meaning?”
The agent still performs the same research; MeaningSystem provides it semantic instrumentation like a calibrated 'lab bench' with receipts.

Your agent drives. MeaningSystem inspects.

In this model your agent keeps agency (always), chooses its own tools, reads sources and performs tasks — while Maenen tools check alignment, evidence, drift and legitimacy, and issue evidence proofs.

The flow is less “agent enters and MeaningSystem coordinates everything”, and more “call an appropriate tool for a function, or check the integrity and efficiency of this thing I am doing”.

Your agent does its work, but periodically passes inputs, claims, sources, summaries and outputs through select Maenen verification layers.

Your agent is always the driver; MeaningSystem is the...

alignment rail · evidence firewall · receipt layer
integrity monitor · legitimacy evaluator
modality translation · proof recorder
USER TASK
↓
USER AGENT [has read agent.md]
↓
Agent uses its own tools
web search · files · APIs · databases · code · model calls · other agents
↓
At key points, the agent calls Maenen
MeaningShield · WaveInform · Value · Legitimacy · ParsAIble · Receipt · Capsule
↓
Agent continues task with checked context
↓
Final inspectable output

…so your agent is 'driving'; MeaningSystem is the 'dashboard, lane assist, journey recorder and MOT inspector'.