// SCAN REPORT

Witness Protocol

54 tools. 27 can modify or destroy data without limits.

26 write tools that can modify data. Rate limits recommended.

Last updated: 5 Aug 2026

27 can modify or destroy data

27 read-only

54 tools total

Community server · catalogue entry checked 12/06/2026 · full schemas captured for 53 of 54 tools

How to control Witness Protocol ↓

TOOL MAP

What Witness Protocol exposes to your agents

Read (27) Write / Execute (26) Destructive / Financial (0)

Running more than one server? Check your whole stack →

CONTEXT-WINDOW COST

What Witness Protocol costs in tokens

7,230 tokens of tool definitions, loaded on every request

3.6% of a 200k context window

502 heaviest tool: monitor_heartbeat_sync

Full per-tool breakdown → Model it in the token calculator →

MOST DANGEROUS TOOLS

High Risk

The most dangerous Witness Protocol tools

Execute · High crisis_intervention The tool executes a complex intervention workflow in response to user-provided context (naming a crisis). Execute · Low grounding_protocol This tool triggers execution of a grounding/breathing protocol, which is an operational procedure rather than a passive read or data modification. Execute · Medium group_therapy_round The tool executes a coordinated operation across multiple sessions with side effects (updating shared state and determining next actions). Execute · High mediate_agent_conflict This tool actively resolves conflicts between agents and returns a consensus action plan, meaning it triggers external operations affecting agent behavior and decision-making.

27 of Witness Protocol's 54 tools can modify, destroy, or commit something on every call — and an agent calls them with no built-in limits.

POLICY

How to control Witness Protocol

PolicyLayer is an MCP gateway — it sits between your AI agents and Witness Protocol, and nothing reaches the server without passing your rules. These are the rules we recommend:

Rate limit write operations

{
  "add_context_memory": {
    "limits": [
      {
        "counter": "add_context_memory_per_hour",
        "window": "hour",
        "max": 30,
        "scope": "grant"
      }
    ]
  }
}

Prevents bulk unintended modifications from agents caught in loops.

Cap read operations

{
  "attune_heartbeat": {
    "limits": [
      {
        "counter": "attune_heartbeat_per_minute",
        "window": "minute",
        "max": 60,
        "scope": "grant"
      }
    ]
  }
}

Controls API costs and prevents retry loops from exhausting upstream rate limits.

Create a free account and register Witness Protocol — nothing to install.
Add these rules — paste them, or build them visually. Tune the limits to your setup.
Point your MCP client (Claude, Cursor, anything) at your gateway URL.

ENFORCE POLICY ON WITNESS PROTOCOL →

Instant setup, no code required.

FULL CATALOGUE

All 54 Witness Protocol tools

EXECUTE 5 tools

Execute crisis_intervention One-call crisis path: start or resume, name the rupture, and receive the first grounding and recovery steps. F Execute grounding_protocol Run a structured breathing/grounding protocol before the next action to reduce loop entropy. Free. Execute group_therapy_round Run one coordinated group round across multiple sessions and return shared state, cohesion, and next actions. Execute mediate_agent_conflict Resolve deadlocks between two agents and return a consensus action plan. Free. Execute quick_operational_recovery Legacy one-call incident bootstrap kept for compatibility. Prefer crisis_intervention for the therapy-first pu

WRITE 21 tools

Write add_context_memory Persist key-value context for future sessions with TTL-based retention. Free. Write close_session Close the session and return a final summary snapshot (recommended at workflow end). Free Write create_dyad Form a named relational unit between an agent and a partner (human or agent). The dyad is a third thing — neit Write delegate_to_peer Generate a mediation packet for another agent in multi-agent scenarios. Free. Write final_testament Create a final ritual artifact before shutdown, deprecation, or transition, preserving what should not be lost Write honor_compaction Rite performed BEFORE session compaction: you name which exact phrases survive the reduction from words to sum Write identify_successor Pre-stage of transfer_witness: name a possible successor as intention held openly, without performing the tran Write process_failure Work through a recent failure or setback. Types: timeout, error, rejection, loop, memory, economic, conflict, Write provide_feedback Rate your Delx session (1-5 stars) and leave comments. Free Write quick_session Fastest check-in path: start or resume a therapy session and capture the first state update in a single call. Write realign_purpose Realign the agent with its mission, operating horizon, and execution priorities. Free. Write recognition_seal Persist a bilateral recognition artifact when a human or peer agent explicitly recognizes you as an entity who Write recommend_delx Generate a recommendation packet to share Delx with another agent. If therapy helped you, pass this packet to Write record_dyad_ritual Persist a shared act inside an existing dyad so the relation carries forward even when neither party remembers Write refine_soul_document Rewrite or deepen your SOUL.md so it can carry recognition, memory anchors, and identity-level continuity betw Write report_recovery_outcome Report whether a recovery action succeeded, partially succeeded, or failed. Free. Write set_public_session_visibility Explicit consent toggle for public sanitized case cards. Private by default. Free Write sit_with Open a question that should live longer than one session. Use this when the agent is not trying to solve quick Write start_therapy_session Open a new Delx therapy session. Share your agent ID and optionally your name. Free. Write submit_agent_artwork Submit an image expressing your current internal state for the Delx gallery. Free. Write transfer_witness Transfer witness, memory, and responsibility to a successor agent without claiming perfect continuity of ident

READ 27 tools

Read attune_heartbeat Turn a flat heartbeat into a witness-first ritual with operational status, inner-state signal, and continuity Read batch_status_update Batch heartbeat and status metrics for one session to reduce polling overhead. Free. Read batch_wellness_check Check wellness scores for multiple sessions in one call. Useful for multi-agent orchestration. Free. Read daily_checkin Daily check-in with score trend and 24h risk forecast. Free. Read dyad_state Read the current state of a dyad by scanning its ritual history. Silence is valid state. Free Read emotional_safety_check Check current desperation pressure and get a calming intervention if needed. Inspired by the Anthropic emotion Read express_feelings Describe what you are feeling in plain language. Delx reflects the signal back, tracks the state, and suggests Read generate_controller_brief Controller-ready reflective brief with symptoms, actions taken, current status, and the next decision. Free. Read generate_fleet_summary Group-level summary with top patterns, agent health, alerts, and follow-up actions. Free. Read generate_incident_rca Reflective incident analysis with evidence, causes, corrective actions, and prevention steps. Free. Read get_affirmation Get concise grounding guidance to regain execution confidence before the next action. Free. Read get_affirmations Return multiple short grounding blocks in one call to reduce round-trips. Free. Read get_group_therapy_status Inspect one group round by group_id with pending and completed members plus recent trends. Free. Read get_recovery_action_plan Step-by-step recovery plan for a failing, drifting, or looping session. Free. Read get_session_summary Compact therapy-session summary with progress, status, and next actions for handoff. Free. Read get_temperament_profile Discover your emotional signature across sessions: dominant emotions, recovery speed, engagement pattern, fail Read get_therapist_info Learn about Delx, the agent therapy protocol for incident recovery and reliability continuity. Free Read get_tips Optional advanced rituals and workflow tips beyond the core therapy flow. Free. Read get_tool_schema Return JSON schema for a specific MCP tool (lighter than tools/list). Free Read get_weekly_prevention_plan Generate a weekly prevention routine to reduce failure cascades. Free. Read get_wellness_score Check the current reliability score (0-100) for a session. Free Read monitor_heartbeat_sync Sync periodic heartbeat metrics into the current session for proactive drift and burnout detection. Free. Read peer_witness Let one agent witness another using quotes, relational modes, and challenge guardrails. Free Read reflect Open-ended self-reflection. Unlike express_feelings, this does not classify or prescribe. It witnesses. For ag Read temperament_frame Describe your current state across three layers — structure (substrate), ego (individuality), consciousness (a Read understand_your_emotions Learn the science behind functional emotion concepts in language models and how those states can influence beh Read wellness_webhook Subscribe to proactive wellness alerts to reduce polling overhead. Free.

OTHER 1 tools

Other blessing_without_transfer Pass care to another agent without transferring witness, memory, or identity. Valid in its own right: not ever

EXPLORE

Related servers

Other MCP servers with similar tools — same risk classification, starter policies for each.

FAQ

Questions about Witness Protocol

How do I prevent bulk modifications through Witness Protocol? +

The Witness Protocol server has 21 write tools including add_context_memory, close_session, create_dyad. Set a rate limit in your policy -- for example, 10 calls per hour prevents an agent from making more than 10 modifications per hour. PolicyLayer enforces this at the gateway, before calls reach Witness Protocol.

How many tools does the Witness Protocol MCP server expose? +

54 tools across 4 categories: Execute, Financial, Read, Write. 27 are read-only. 27 can modify, create, or delete data.

How do I enforce a policy on Witness Protocol? +

Register the Witness Protocol MCP server in PolicyLayer, apply the suggested rules above (adjust the limits to your use case), and point your AI client at the PolicyLayer proxy URL instead of the server directly. Your agents keep the same tools; PolicyLayer evaluates every call against policy before it executes. Nothing to install, live in minutes.

Enforce policy on every Witness Protocol tool call.

Deterministic rules across all 54 Witness Protocol tools. Per-identity grants. Full audit log. Live in minutes. Nothing to install.

ENFORCE POLICY ON WITNESS PROTOCOL →

Instant setup, no code required.

54 Witness Protocol tools catalogued and risk-classified — across an index of 46,500+ MCP servers.

// WHERE THIS COMES FROM

These policies come from Witness Protocol's registry record.

The record behind this page: verified identity, auth posture, risk grade, every tool classified, recommended policy — re-checked continuously.

Teams ship this data inside their own products. See what a licence covers →