Build with Rippletide.

Add a decision call to your existing agent. Keep rules and corrections in one decision system, across calls and agents.

Start hereMake your first Decision API call

Decision API / Quickstart

Make your first decision.

Use a Connection key from Decision API in the platform. Keep it in your backend environment.

Terminalexport RIPPLETIDE_API_KEY='your-connection-key'

Run the example in your backend. Keep the returned decision_id to attach feedback to that decision.

Understand the result

A choice your agent can use.

The response keeps the chosen action, scores and supporting context together.

choice

The selected option from the choices you supplied.

choice_scores

The model’s score for each candidate action.

evidence

The memory excerpts and policy versions supplied for this decision.

decision_id

The identifier to use when sending a correction.

Specialize the decisions

Your rules, across calls.

Save an active policy in the same decision bank. Subsequent calls can use that policy without adding its full text to every request.

PUT/v1/policies/:policyId

Example policy

For a package marked delivered but not received, request carrier delivery evidence before recommending a replacement or refund.

Download executable policy example

Feedback

Give the next decision more to work with.

Send the original decision ID, the correction and its source. The Decision API stores that correction in the same bank for later retrieval.

POST/v1/feedback

Download the Python feedback example

Portable decisions

New agent. Same decision system.

Connect your next backend agent to the same bank through the API. Its policies, decision history and retained corrections stay available when the calling agent or its LLM changes.

Agent AGPT-based agent

Same ConnectionPolicies + history + corrections

Agent BClaude-based agent

Keep the Connection key in the backend. The bank is selected by that key.

Tide-1 Coming soon

Specialize. Correct. Improve.

Tide-1 will train specialized decision models from your rules and corrections. Keep the model when you change agents or LLMs.

Request early access

Tide-Enterprise

Company data.
One decision reference.

Build a shared hypergraph from the information your agents need. Follow the same three stages as the platform.

01 / CONNECT

Connect your data

Bring documents, policies, business records and decision history into the hypergraph.

02 / CHECK

Check your data

Compare CRM and ERP identifiers, an approved policy with a conflicting help-center page, and missing fields in operational APIs. Reconcile the source data into a common decision reference.

03 / INTEGRATE

Integrate & enforce

Check each proposed action against validated rules before its tool executes. Enforce the decision at runtime with deterministic execution, explanations and traceability.

Runtime reinforcement brings outcomes and reviewed corrections back into the reference, with replay before release.

Contact Rippletide