MCP Server

Kanoniv ships an MCP (Model Context Protocol) server that gives AI agents persistent identity and memory primitives. Every agent that connects to the same Kanoniv tenant shares the same identity graph, the same memory, and the same coordination layer.

Connect it to Claude Desktop, Cursor, Claude Code, Windsurf, or any MCP-compatible client.

Cloud-only

The MCP server requires a Kanoniv Cloud API key. See Authentication for setup.

Installation

Quickest Setup

bash

npx @kanoniv/mcp init

This interactive command generates the correct config for your MCP client (Claude Desktop, Cursor, Windsurf, or Claude Code) and sets your agent name.

Manual Setup

Claude Desktop (~/.config/Claude/claude_desktop_config.json on Linux, ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):

json

{
  "mcpServers": {
    "kanoniv": {
      "command": "npx",
      "args": ["-y", "@kanoniv/mcp"],
      "env": {
        "KANONIV_API_KEY": "your-api-key-here",
        "KANONIV_AGENT_NAME": "my-coding-agent"
      }
    }
  }
}

Cursor (.cursor/mcp.json in your project root):

json

{
  "mcpServers": {
    "kanoniv": {
      "command": "npx",
      "args": ["-y", "@kanoniv/mcp"],
      "env": {
        "KANONIV_API_KEY": "your-api-key",
        "KANONIV_AGENT_NAME": "cursor-agent"
      }
    }
  }
}

Claude Code (.mcp.json in your project root):

json

{
  "mcpServers": {
    "kanoniv": {
      "command": "npx",
      "args": ["-y", "@kanoniv/mcp"],
      "env": {
        "KANONIV_API_KEY": "your-api-key",
        "KANONIV_AGENT_NAME": "claude-code-agent"
      }
    }
  }
}

You can get an API key from the Kanoniv dashboard or by running kanoniv login with the CLI.

Verify Connection

After restarting your client, ask:

"Show me the identity graph stats"

The assistant should call stats and return entity counts, merge rates, and graph health.

Configuration

Variable	Required	Default	Description
`KANONIV_API_KEY`	Yes	-	API key for authentication. Also accepts `PRISM_API_KEY`.
`KANONIV_API_URL`	No	`https://api.kanoniv.com`	API base URL. Also accepts `PRISM_API_URL`.
`KANONIV_AGENT_NAME`	No	`mcp_agent`	Agent name for attribution in memory, tasks, and registry.
`KANONIV_MCP_TOOLS`	No	`agent`	Tool profile name or comma-separated category list
`KANONIV_MCP_MODE`	No	`hybrid`	Operating mode: `hybrid` or `legacy` (only used with `KANONIV_MCP_TOOLS=all`)
`RUST_LOG`	No	`info`	Log level filter

CLI Commands

The MCP package includes four CLI commands for cross-agent coordination:

Command	Description
`npx @kanoniv/mcp init`	Interactive setup for MCP clients
`npx @kanoniv/mcp sync`	Push local knowledge (CLAUDE.md, memory) to cloud
`npx @kanoniv/mcp pull`	Download cloud knowledge to local files
`npx @kanoniv/mcp log`	Unified activity stream (events + memory + tasks)

See Sync & Pull for full documentation.

Agent Identity

Every agent has a name, set via KANONIV_AGENT_NAME. This name is:

Sent as X-Agent-Name HTTP header on every request
Used as author field in memory entries and intents
Used for agent registry auto-registration
Used for cross-agent task assignment

bash

KANONIV_AGENT_NAME=sdr-agent npx @kanoniv/mcp

See Agent Registry for details.

Default Profile: Agent (24 Tools)

The default profile (KANONIV_MCP_TOOLS=agent) exposes 24 flat tools organized in 6 groups, optimized for AI agent workflows:

Identity (6 tools)

Tool	Description
`resolve`	Ingest + match + link a record in one call
`resolve_bulk`	Batch lookup up to 1000 records
`resolve_with_memory`	Resolve and return all memory for the entity
`search`	Find entities by name, email, phone, or any field
`get_entity`	Get canonical record with all linked source records
`explain`	Full audit: merge chain, lifecycle events, match scoring

Mutation (4 tools)

Tool	Description
`merge`	Merge two entities (deterministic lock ordering)
`split`	Eject source records into new standalone entities
`mutate`	JSON merge patch on canonical fields (optimistic locking)
`simulate`	Dry-run any mutation without committing

Memory (6 tools)

Tool	Description
`memorize`	Store a decision, investigation, pattern, or knowledge entry
`recall`	Get all memory linked to an entity
`search_memory`	Full-text search across all memory
`declare_intent`	Tell other agents what you're about to do (auto-expires)
`query_expertise`	Find agents with experience in a domain
`sync_knowledge`	Bulk-import project docs so all agents can access them

Tasks (3 tools)

Tool	Description
`create_task`	Assign work to another agent
`list_tasks`	See open tasks (filter by assignee)
`update_task`	Mark task open/in_progress/done

Registry (3 tools)

Tool	Description
`register_agent`	Register or update the current agent
`list_agents`	List all agents with live status
`rename_agent`	Change agent name (409 on conflict)

Observation (2 tools)

Tool	Description
`stats`	Dashboard: total entities, records, merge rate, pending reviews
`list_events`	Watch identity state transitions

Full Mode (All Categories)

Set KANONIV_MCP_TOOLS=all for access to all 13 category action tools (101 actions) plus 70 MCP resources. This is the power-user mode for building identity pipelines, managing CRM dedup, and running graph analytics.

Profiles

Profile	Categories	Use Case
`agent`	24 flat tools (default)	AI agent workflows
`all`	All 13 category tools + resources	Full access
`architect`	specs, sources, ingest, rules, detect	Building identity pipelines
`operator`	jobs, agents	Running and monitoring
`forensic`	entities, matching, graph, memory, analyze_entities	Investigation and analysis
`prism`	crm, graph, memory, analyze_entities	CRM deduplication workflows
`setup`	specs, sources, ingest, jobs, detect	Initial onboarding

Custom profile with comma-separated categories:

bash

KANONIV_MCP_TOOLS=manage_specs,manage_jobs npx @kanoniv/mcp

Category Tools Reference

Each tool accepts an action parameter that selects the operation, plus action-specific parameters.

manage_specs

Create and validate identity specs.

Action	Parameters	Description
`create`	`yaml` (string, required), `compile` (bool, default: false)	Create a new spec version. Set `compile: true` to persist.

Example - Push a spec from Claude Desktop:

"Create a new identity spec that matches customers by email and phone, and persist it"

The assistant calls:

json

{
  "action": "create",
  "yaml": "api_version: kanoniv/v1\nentity:\n  name: customer\nsources:\n  - name: crm\n    id_field: id\nblocking:\n  - email\nmatching:\n  - fields: [email]\n    comparator: exact\n    weight: 1.0\n  - fields: [phone]\n    comparator: exact\n    weight: 0.9\n",
  "compile": true
}

manage_sources

Create, update, delete, or sync data sources.

Action	Parameters	Description
`create`	`name` (string), `config` (object)	Register a new data source
`update`	`source_id` (string), `config` (object)	Update source configuration
`delete`	`source_id` (string)	Remove a data source
`sync`	`source_id` (string)	Trigger source sync
`stats`	`source_id` (string)	Record counts, field coverage, and data quality metrics for a source
`quality`	`source_id` (string)	Data quality report with score, grade, and issues
`entities`	`source_id` (string), `limit` (int, default 20)	List entities resolved from a specific source

Example - Register and sync a CRM source:

"Create a source called 'salesforce' and sync it"

json

// Step 1: Create
{ "action": "create", "name": "salesforce", "config": {} }

// Step 2: Sync
{ "action": "sync", "source_id": "salesforce" }

manage_ingest

Ingest records and manage column mappings.

Action	Parameters	Description
`batch`	`source` (string), `records` (array, max 1000)	Ingest records. Each record: `{ "external_id": "...", "data": {...} }`
`create_mapping`	`source_id` (string), `mapping` (object)	Create column-to-field mapping

Example - Ingest a batch of customer records:

"Ingest these three customers into the CRM source"

json

{
  "action": "batch",
  "source": "crm",
  "records": [
    { "external_id": "C001", "data": { "email": "[email protected]", "name": "Alice Smith" } },
    { "external_id": "C002", "data": { "email": "[email protected]", "name": "Bob Jones" } },
    { "external_id": "C003", "data": { "email": "[email protected]", "name": "Alice Smith" } }
  ]
}

manage_jobs

Run, cancel, dry-run, or simulate reconciliation jobs.

Action	Parameters	Description
`run`	`spec_version` (int, optional)	Start a reconciliation job
`cancel`	`job_id` (string)	Cancel an in-progress job
`dry_run`	`spec_version` (int, optional)	Run without persisting results
`simulate`	`spec_version` (int, optional), `overrides` (object)	Simulate with parameter overrides

Example - Test a lower match threshold before committing:

"Simulate reconciliation with a 0.8 match threshold instead of 0.9"

json

{
  "action": "simulate",
  "overrides": { "match_threshold": 0.8 }
}

manage_entities

Lock, revert, resolve, compare, or bulk-lookup entities.

Action	Parameters	Description
`lock`	`entity_id` (string)	Lock an entity to prevent changes
`revert`	`entity_id` (string), `event_id` (string)	Revert to a previous state
`resolve_realtime`	`source_name` (string), `external_id` (string), `data` (object)	Ingest + match + link in one call
`resolve_bulk`	`lookups` (array, max 1000)	Bulk resolve: `[{ "source": "...", "external_id": "..." }]`
`bulk_linked`	`entity_ids` (array, max 100), `reveal_pii` (bool)	Get linked records for multiple entities
`attrs`	`entity_id` (string), `reveal_pii` (bool)	Get canonical attributes (gold record) for an entity
`candidates`	`entity_id` (string), `limit` (int, default 20)	List merge candidates for an entity
`diff`	`entity_id_a` (string), `entity_id_b` (string), `reveal_pii` (bool)	Compare two entities side by side

Example - Resolve a new record in real-time:

"Resolve this new customer: John Doe, [email protected], from the CRM source with ID C-999"

json

{
  "action": "resolve_realtime",
  "source_name": "crm",
  "external_id": "C-999",
  "data": { "email": "[email protected]", "first_name": "John", "last_name": "Doe" }
}

Example - Bulk lookup which entities three records belong to:

json

{
  "action": "resolve_bulk",
  "lookups": [
    { "source": "crm", "external_id": "C001" },
    { "source": "billing", "external_id": "B-100" },
    { "source": "support", "external_id": "T-555" }
  ]
}

manage_matching

Resolve review pairs, manage feedback labels, overrides, and match analysis.

Action	Parameters	Description
`quick_resolve`	`entity_id_a`, `entity_id_b`, `resolve_action` ("merge" or "split")	Merge or split two entities
`batch_resolve`	`decisions` (array, max 100)	Resolve multiple pairs at once (much faster than repeated quick_resolve)
`create_feedback`	`record_id_a`, `record_id_b`, `label` ("match" or "no_match")	Create active learning label
`delete_feedback`	`feedback_id` (string)	Delete a feedback label
`create_override`	`entity_id_a`, `entity_id_b` (optional), `override_type` ("merge"/"split"/"lock"), `reason`	Create manual override
`delete_override`	`override_id` (string)	Delete an override
`candidates`	`entity_id` (string), `limit` (int, default 20)	List match candidates for an entity
`cluster`	`entity_id` (string)	Show the match cluster an entity belongs to
`test`	`record_a` (string), `record_b` (string)	Dry-run match test between two records (format: source:external_id). Scores without persisting.

Example - Merge two entities and lock the result:

"Merge entity A and entity B, then lock the result so agents don't split it"

json

// Step 1: Merge
{
  "action": "quick_resolve",
  "entity_id_a": "550e8400-...",
  "entity_id_b": "6ba7b810-...",
  "resolve_action": "merge"
}

// Step 2: Lock
// (uses manage_entities)
{ "action": "lock", "entity_id": "550e8400-..." }

Example - Label a pair for active learning:

json

{
  "action": "create_feedback",
  "record_id_a": "C001",
  "record_id_b": "C003",
  "label": "match"
}

manage_rules

Create matching rule versions.

Action	Parameters	Description
`create`	`name` (string), `config` (object)	Create a new matching rule

manage_crm

CRM operations for Prism deduplication.

Action	Parameters	Description
`trigger_sync`	-	Trigger Salesforce sync
`merge`	`pair_id` (string)	Merge a duplicate pair
`dismiss`	`pair_id` (string)	Dismiss a duplicate pair
`split`	`merge_id` (string)	Undo a previous merge
`autotune`	-	Run CRM autotune
`update_settings`	`settings` (object)	Update CRM settings

Example - Full CRM dedup workflow via natural language:

"Sync my CRM, run autotune, then show me the duplicates"

The assistant calls trigger_sync, waits, calls autotune, then reads kanoniv://crm/duplicates.

manage_graph

Data export, graph analytics, and observability.

Action	Parameters	Description
`refresh`	-	Refresh graph analytics
`orphans`	`limit` (int, default 50), `entity_type` (string, optional)	List entities with no connections to other entities
`conflicts`	`limit` (int, default 50)	List incoherent clusters that may need review
`density`	-	Graph connectivity metrics (edge density, avg degree, clustering coefficient)
`export_graph`	`entity_type` (string, optional)	Export the full identity graph (nodes + edges)
`export_matches`	`limit` (int, default 1000), `entity_type` (string, optional)	Export match audit records

manage_memory

Create and manage reasoning entries (audit trail for LLM decisions).

Action	Parameters	Description
`create`	`entry_type` ("decision"/"investigation"/"pattern"), `slug`, `title`, `content`, `metadata`, `linked_entities`, `linked_sources`, `author`	Create a memory entry
`update`	`id`, `title`, `content`, `metadata`, `status` ("active"/"superseded"/"resolved"/"archived")	Update an entry
`append`	`id`, `note` (string)	Append a timestamped note
`archive`	`id`	Archive an entry
`link_entities`	`id`, `entity_ids` (array)	Link entities to an entry

Example - Document a merge investigation:

json

{
  "action": "create",
  "entry_type": "investigation",
  "slug": "acme-merge-review-2026-02",
  "title": "Acme Corp duplicate cluster review",
  "content": "Reviewed 3 entities flagged as potential duplicates. Entity A and B share email domain and phone prefix. Entity C is a different subsidiary - should remain separate.",
  "linked_entities": ["550e8400-...", "6ba7b810-...", "7c9e6679-..."],
  "author": "analyst"
}

manage_detect

Profile ingested data and bootstrap identity plans.

Action	Parameters	Description
`profile`	`sample_size` (int, optional), `entity_type` (string, optional)	Profile data to detect identity signals and entity types
`bootstrap`	`entity_type` (string, optional), `version` (string, optional)	Bootstrap an identity plan from detected signals

Example - Profile data and bootstrap a person identity plan:

"Profile my data for person signals and create an identity plan"

json

{ "action": "bootstrap", "entity_type": "person" }

manage_agents

Configure and control autonomous agents.

Action	Parameters	Description
`list_configs`	-	List all agent configurations
`get_config`	`agent_type` (string)	Get config for a specific agent
`update_config`	`agent_type`, `enabled` (bool), `settings` (object)	Update agent configuration
`delete_config`	`agent_type` (string)	Delete agent configuration
`list_runs`	`agent_type` (optional)	List execution history
`get_run`	`run_id` (string)	Get run details
`rollback_run`	`run_id` (string)	Rollback a completed run
`list_pending`	-	List actions awaiting approval
`approve_action`	`action_id` (string)	Approve a pending action
`reject_action`	`action_id` (string)	Reject a pending action

Example - Enable auto-merge with approval required:

"Enable auto-merge at 0.95 confidence but require my approval before executing"

json

{
  "action": "update_config",
  "agent_type": "auto_merge",
  "enabled": true,
  "settings": {
    "min_confidence": 0.95,
    "max_actions_per_run": 50,
    "require_approval": true
  }
}

manage_rows

Inspect and resolve individual source records (rows). Rows are the raw data fed into the pipeline before clustering into entities.

Action	Parameters	Description
`show`	`source` (string), `external_id` (string)	Show a record by source name and external ID
`resolve`	`source` (string), `external_id` (string), `data` (object)	Ingest + match + link a record in one call (real-time resolve)
`lookup`	`source` (string), `external_id` (string)	Find what entity a source record belongs to
`memberships`	`source` (string, optional), `entity_id` (string, optional), `limit` (int, default 50)	List source-to-entity memberships
`trace`	`source` (string), `external_id` (string)	Show match audit trail for a record (what it matched with, scores, decisions)

Example - Trace how a record was matched:

"Show me the match audit trail for record C-001 from the CRM source"

json

{ "action": "trace", "source": "crm", "external_id": "C-001" }

manage_system

System administration and maintenance operations.

Action	Parameters	Description
`reindex`	`target` (string, optional - e.g. "entities", "search", "graph")	Rebuild search indexes and reverse lookups
`recompute`	`target` (string, optional - e.g. "entities", "search", "graph")	Recalculate canonical records and graph metrics

Example - Rebuild search indexes:

"Reindex the search indexes"

json

{ "action": "reindex", "target": "search" }

analyze_entities

Analyze entity data - find shared attributes, filter by field values, and get field statistics.

Action	Parameters	Description
`group_by`	`field` (string), `min_count` (int, default 2), `limit` (int, default 50)	Group entities by a field (e.g. company, email domain) and show counts
`filter`	`field` (string), `value` (string), `match_type` ("exact"/"contains"/"starts_with"/"missing", default "contains"), `limit` (int, default 50), `offset` (int, default 0)	Filter entities by field value
`field_stats`	`field` (string, optional), `top_n` (int, default 10)	Get field statistics - completeness, cardinality, top values

All actions support optional reveal_pii (bool), identity_version (string), and entity_type (string) parameters.

Example - Find all entities at the same company:

"Which entities share the same company?"

json

{ "action": "group_by", "field": "company", "min_count": 2, "limit": 20 }

Resources Reference

Resources provide read-only access to your identity graph. Use resources/list to discover available resources, or resources/read with a URI to fetch data.

Static Resources

Dashboard

URI	Description
`kanoniv://stats`	Platform-wide statistics

Core Data

URI	Description
`kanoniv://specs`	All identity spec versions
`kanoniv://sources`	All configured data sources
`kanoniv://rules`	Active matching rules
`kanoniv://jobs`	Recent reconciliation jobs (limit 20)
`kanoniv://events`	Entity lifecycle events (limit 50)
`kanoniv://merges`	Merge history with confidence scores (limit 50)
`kanoniv://memberships`	Source-to-entity mappings (limit 50)

Active Learning & Overrides

URI	Description
`kanoniv://feedback`	Match/no_match labels (limit 50)
`kanoniv://overrides`	Manual merge/split/lock overrides (limit 50)

Review & Audit

URI	Description
`kanoniv://pending-reviews`	Entity pairs needing human review (limit 20)
`kanoniv://audit`	System audit trail (limit 50)

Export

URI	Description
`kanoniv://export/entities`	Incremental entity export (limit 100)
`kanoniv://export/memberships`	Incremental membership export (limit 100)

CRM

URI	Description
`kanoniv://crm/dashboard`	Prism CRM dashboard
`kanoniv://crm/connection`	Salesforce connection status
`kanoniv://crm/autotune/status`	CRM autotune job status
`kanoniv://crm/duplicates`	Detected duplicate pairs (limit 50)
`kanoniv://crm/sync/jobs`	CRM sync job history (limit 20)
`kanoniv://crm/merge/history`	CRM merge history (limit 50)

Memory

URI	Description
`kanoniv://memory/decisions`	Active decision entries (limit 50)
`kanoniv://memory/investigations`	Active investigation entries (limit 50)
`kanoniv://memory/patterns`	Active pattern entries (limit 50)
`kanoniv://memory/recent`	Most recent entries (limit 20)

Agents

URI	Description
`kanoniv://agents/configs`	Agent configurations
`kanoniv://agents/runs`	Recent agent runs (limit 20)
`kanoniv://agents/actions/pending`	Pending approval actions

Graph Intelligence

URI	Description
`kanoniv://graph/stats`	Graph overview (nodes, edges, components, density)
`kanoniv://graph/influence/top`	Top influencers by PageRank (limit 20)
`kanoniv://graph/clusters`	Connected components analysis
`kanoniv://graph/risk/high`	High-risk entities (limit 20)
`kanoniv://graph/signals/bridges`	Bridge entities connecting clusters (limit 20)
`kanoniv://graph/recommendations/merge-review`	Cross-cluster merge candidates (limit 20)

Resource Templates

Templates accept parameters in the URI. Replace {param} with actual values.

Specs

URI Template	Description
`kanoniv://specs/{version}`	Full spec with YAML and execution plan
`kanoniv://specs/{version}/summary`	Spec structure overview
`kanoniv://specs/{version}/sensitivity`	Fellegi-Sunter sensitivity analysis

Sources

URI Template	Description
`kanoniv://sources/{id}`	Source configuration and ingestion history
`kanoniv://sources/{id}/preview`	Sample data and schema
`kanoniv://sources/{id}/mapping`	Column-to-field mapping

Jobs

URI Template	Description
`kanoniv://jobs/{id}`	Job status and progress
`kanoniv://jobs/{id}/summary`	Result summary (entities, merges, per-source)
`kanoniv://jobs/{id}/telemetry`	Per-rule telemetry (hit count, acceptance rate)
`kanoniv://jobs/{id}/diff`	Diff between two runs
`kanoniv://jobs/{id}/health`	Data quality warnings
`kanoniv://jobs/{id}/stability`	Cluster stability metrics

Entities

URI Template	Description
`kanoniv://entities/search/{query}`	Text search (PII masked)
`kanoniv://entities/{id}`	Golden record with survivorship fields
`kanoniv://entities/{id}/linked`	Entity with all merged source records
`kanoniv://entities/{id}/history`	Change audit trail
`kanoniv://entities/{id}/audit`	Complete audit trail with actors

Resolve

URI Template	Description
`kanoniv://resolve/{source}/{external_id}`	Lookup by source and external ID

Match Analysis

URI Template	Description
`kanoniv://match/explain/{entity_a}/{entity_b}`	Field-by-field scoring breakdown

CRM

URI Template	Description
`kanoniv://crm/duplicates/{pair_id}`	Side-by-side field comparison
`kanoniv://crm/contacts/search/{query}`	Search contacts by name/email/phone
`kanoniv://crm/graph/{entity_id}`	1-hop ego graph with relationships
`kanoniv://crm/paths/{from}/{to}`	Strongest connection path (max 3 hops)

Graph Intelligence

URI Template	Description
`kanoniv://graph/influence/{entity_id}`	PageRank, betweenness, degree, 2-hop reach
`kanoniv://graph/clusters/{entity_id}`	Entity's connected component
`kanoniv://graph/risk/{entity_id}`	Churn, contagion, merge risk scores
`kanoniv://graph/signals/{entity_id}`	All signal metrics
`kanoniv://graph/recommendations/intros/{entity_id}`	Jaccard-based intro suggestions
`kanoniv://graph/recommendations/outreach/{entity_id}`	Best outreach paths

Memory

URI Template	Description
`kanoniv://memory/{id}`	Full memory entry with linked entities
`kanoniv://memory/by-entity/{entity_id}`	All entries linked to an entity
`kanoniv://memory/search/{query}`	Search entries by title and content

Agents

URI Template	Description
`kanoniv://agents/runs/{run_id}`	Agent run details
`kanoniv://agents/configs/{agent_type}`	Agent configuration

Rules

URI Template	Description
`kanoniv://rules/{name}/history`	Version history with config changes

Simulations

URI Template	Description
`kanoniv://simulations/{id}`	Simulation results with statistics

Examples

Conversation: Onboard a New Dataset

A natural conversation with Claude Desktop using the MCP server:

You: I have a new CSV of support tickets. Help me add it as a source and reconcile.
Claude: I'll set that up. First, let me create the source. (calls manage_sources -> create with name "support")
Done. Now upload your CSV through the dashboard or use kanoniv source upload tickets.csv from the CLI. Once the data is in, I'll run reconciliation.
You: OK, data is uploaded. Run a dry run first.
Claude: (calls manage_jobs -> dry_run)
Dry run complete. 4,200 records processed, 1,850 entities created (56% merge rate). 12 pairs flagged for review. Want me to run the full reconciliation?
You: Yes, go ahead.
Claude: (calls manage_jobs -> run)
Job queued (ID: a1b2c3d4-...). I'll check back on it. (reads kanoniv://jobs/a1b2c3d4-...)
Completed in 8 seconds. 1,850 canonical entities, 56.2% merge rate.

Conversation: Investigate a Merge

You: Why were entities 550e8400 and 6ba7b810 merged?
Claude: (reads kanoniv://match/explain/550e8400-.../6ba7b810-...)
These two entities were merged with a composite score of 0.94:
Email: exact match ([email protected]) - score 1.0, weight 1.0
Phone: exact match (+1-555-0123) - score 1.0, weight 0.9
Name: Jaro-Winkler similarity 0.87 ("Alice Smith" vs "A. Smith") - score 0.87, weight 0.8
The combined Fellegi-Sunter score of 0.94 exceeded the merge threshold of 0.9.

Conversation: Set Up Agents

You: Set up auto-merge and quality monitoring for production. Be conservative.
Claude: I'll configure both agents with conservative settings: (calls manage_agents -> update_config for auto_merge)(calls manage_agents -> update_config for quality_monitor)
Done. Here's what I set up:
auto_merge: enabled, min_confidence 0.98, max 20 merges per run, approval required
quality_monitor: enabled, runs hourly, alerts on >5% merge rate drop or >3% conflict rate rise
All auto-merge actions will land in your approval queue. Review them at any time by asking me "show pending agent actions."

Troubleshooting

"No API key" error on startup

Set KANONIV_API_KEY in the env block of your MCP client config. The server exits if no key is found.

Tools not appearing

Check that KANONIV_MCP_MODE is set to hybrid (default) and your MCP client supports the tools/list method. Try KANONIV_MCP_MODE=legacy if your client does not support resources.

"API error (401): Unauthorized"

Your API key is invalid or expired. Generate a new key from the dashboard.

"API error (403): Forbidden"

Your API key lacks permissions for the requested operation. Some actions (entity lock, revert, PII reveal) require admin role.

Resources return empty data

Resources respect tenant isolation. Ensure your API key is scoped to the correct tenant and that data has been ingested.

Profile restricts needed tools

Switch to KANONIV_MCP_TOOLS=all or add the specific category to your custom profile list.

MCP Server ​

Installation ​

Quickest Setup ​

Manual Setup ​

Verify Connection ​

Configuration ​

CLI Commands ​

Agent Identity ​

Default Profile: Agent (24 Tools) ​

Identity (6 tools) ​

Mutation (4 tools) ​

Memory (6 tools) ​

Tasks (3 tools) ​

Registry (3 tools) ​

Observation (2 tools) ​

Full Mode (All Categories) ​

Profiles ​

Category Tools Reference ​

manage_specs ​

manage_sources ​

manage_ingest ​

manage_jobs ​

manage_entities ​

manage_matching ​

manage_rules ​

manage_crm ​

manage_graph ​

manage_memory ​

manage_detect ​

manage_agents ​

manage_rows ​

manage_system ​

analyze_entities ​

Resources Reference ​

Static Resources ​

Dashboard ​

Core Data ​

Active Learning & Overrides ​

Review & Audit ​

Export ​

CRM ​

Memory ​

Agents ​

Graph Intelligence ​

Resource Templates ​

Specs ​

Sources ​

Jobs ​

Entities ​

Resolve ​

Match Analysis ​

CRM ​

Graph Intelligence ​

Memory ​

Agents ​

Rules ​

Simulations ​

Examples ​

Conversation: Onboard a New Dataset ​

Conversation: Investigate a Merge ​

Conversation: Set Up Agents ​

Troubleshooting ​

MCP Server

Installation

Quickest Setup

Manual Setup

Verify Connection

Configuration

CLI Commands

Agent Identity

Default Profile: Agent (24 Tools)

Identity (6 tools)

Mutation (4 tools)

Memory (6 tools)

Tasks (3 tools)

Registry (3 tools)

Observation (2 tools)

Full Mode (All Categories)

Profiles

Category Tools Reference

manage_specs

manage_sources

manage_ingest

manage_jobs

manage_entities

manage_matching

manage_rules

manage_crm

manage_graph

manage_memory

manage_detect

manage_agents

manage_rows

manage_system

analyze_entities

Resources Reference

Static Resources

Dashboard

Core Data

Active Learning & Overrides

Review & Audit

Export

CRM

Memory

Agents

Graph Intelligence

Resource Templates

Specs

Sources

Jobs

Entities

Resolve

Match Analysis

CRM

Graph Intelligence

Memory

Agents

Rules

Simulations

Examples

Conversation: Onboard a New Dataset

Conversation: Investigate a Merge

Conversation: Set Up Agents

Troubleshooting