Outcome verification for agent actions

Verify that agent actions actually completed

Every plan includes preflight validation, policy enforcement, execution tracking, and outcome verification.

Outcome verification

Confirm the real-world result, not the response code

Preflight validation

Catch schema errors before they reach production

Policy controls

Block, require, escalate on risk — per contract

Run evidence

Per-action audit trail with explicit states

Free

Verify your first agent actions. No credit card required.

$0/mo

500 verified actions/month

500 verified actions/month
3 action contracts
1 integration
Preflight schema validation
Policy enforcement (require, block, allow)
False completion detection
MCP server access

Pro

For teams running agents against production APIs.

$79/mo

25,000 verified actions/month

25,000 verified actions/month
Unlimited action contracts
All integration types (Stripe, Salesforce, Zendesk, SendGrid)
Human approval escalation policy
Risk threshold policy rules
Drift event recording
3 projects
MCP server access

Team

For organizations with multiple agent systems and environments.

$299/mo

250,000 verified actions/month

250,000 verified actions/month
Unlimited action contracts
All integration types
All policy rule types
Drift event recording
10 projects
3 environments per project
MCP server access

Business

High-volume verified actions with priority support.

$2000/mo

2,000,000 verified actions/month

2,000,000 verified actions/month
Unlimited projects
Unlimited environments
Unlimited action contracts
All integration types
All policy rule types
Priority support
MCP server access

How verification works

A complete 4-phase action lifecycle. Every phase produces explicit machine-readable state — no ambiguous outcomes.

Preflight

Validate input against the action contract schema. Evaluate all applicable policies. Get an explicit approve, block, or escalate decision.

Policy check

Require fields, block on conditions, enforce risk thresholds, trigger human approval. All decisions logged with machine-readable reason codes.

Execution report

After executing the real-world action, report the result. Distinguish: executing, accepted, completed, failed, unknown. Accepted is not completed.

Verification

Verify the outcome against the real external system. Detect false completions where agents claim success but nothing happened.

Included on every plan: false completion detection

When an agent reports success but the external system has no record of the action, DriftGentic sets final_outcome: false_complete and creates a drift event. The agent receives an explicit warning not to proceed.

×Stripe charge reported succeeded — no charge in Stripe

×CRM record claimed created — no record in Salesforce

×Ticket reported opened — no ticket in Zendesk

×Email claimed sent — no activity in SendGrid

MCP server included on all plans

Connect any MCP-compatible agent framework. All 8 safety tools available automatically via JSON-RPC 2.0.

preflight_validate_action

execute_with_guardrails

report_execution_result

verify_action_outcome

get_run_status

get_action_policy

fetch_contract

check_tool_health

GET $SUPABASE_URL/functions/v1/mcp-server
Authorization: Bearer $API_KEY

Start verifying in minutes

Sign up, define an action contract, and verify your first agent action. The free plan includes everything you need to get started.