Praxa use cases

Praxa exposes several deliberately separate developer surfaces. Start with the job you need to complete, then choose the contract whose authority and release status match that job.

Do not mix credentials between surfaces. Execution Fabric uses personal workspace API keys at api.praxa.io. The Integration Gateway uses deployment-issued, short-lived OAuth tokens. Memory adapters receive clients that your backend owns.

Choose by job

You want to	Example	Use	Availability
Start one durable task and receive a run ID	Submit an incident summary for asynchronous processing	`POST /v1/execute`	Production partner preview; personal tenants
Read, stream, or cancel a durable task	Reconnect an operations dashboard after a network interruption	Runs and events	Production partner preview
Receive run changes at your own HTTPS endpoint	Deliver `run.completed` to a Next.js or Cloudflare Worker webhook	Webhooks	Production partner preview
Read tenant-scoped usage evidence	Show daily requests, tokens, model calls, and billable amount	Usage	Production partner preview
Record natural-language intent for compilation	Ask a deployment to prepare a weekly review without claiming execution	`PraxaClient.submitIntent`	Requires an Integration Gateway deployment
Create and supervise a governed mission	Supply a goal specification, resource budget, signals, and cancellation	Multi-step workflows	Requires an Integration Gateway deployment
Discover policy-eligible capabilities	Search for a calendar-read or code-execution capability before creating a mission	Capability search	Requires an Integration Gateway deployment
Add governed tools to an agent framework	Give an OpenAI, Vercel AI SDK, or LangChain agent the 12 Praxa MCP-compatible tools	MCP integration	Public package; runtime calls require a Gateway deployment
Inspect missions and governance evidence	Read mission events, redacted traces, skills, goals, certificates, and coverage	Monitoring and observability	Requires an Integration Gateway deployment
Query an existing memory system without migrating it	Recall preferences from Mem0 and documents from LangGraph in one request	Memory federation	Public `@praxa/sdk/memory` package
Recall graph entities and relationships	Retrieve Zep edges or Graphiti facts, entities, episodes, and edges	Graph memory tutorial	Public `@praxa/sdk/memory` package
Stage portable memory records in Praxa	Create, lexically query, export, and erase subject-scoped candidates	Hosted memory API	Deployed qualification preview; authenticated canaries pending
Evaluate Praxa evidence without treating it as production proof	Reproduce package checks or inspect the Terminal-Bench pilot protocol	Benchmarks and research	Public evidence artifacts
Add Praxa to an existing agent stack	Use Vercel AI SDK, OpenAI Agents, LangGraph, or an MCP host without replacing the model loop	Agent frameworks	Depends on the selected tool or memory lane
Build a supervised, budgeted agent backend	Combine scopes, approval, events, and independent readback	Governed agent backend	Integration Gateway deployment
Drive an asynchronous application workflow	Use SSE for progress and signed webhooks for durable delivery	Event-driven automation	API key or webhook secret by operation
Discover Praxa product outcomes	Browse all live and beta product capabilities without inferring undocumented APIs	Product capabilities	Praxa product account; developer contracts where listed

Apply the contracts to a product

Product pattern	Concrete example	Build guide	End-to-end tutorial
Product backend	Submit support-incident analysis that survives a browser disconnect	Agent builders	Durable tasks
Coding or desktop handoff	Start a cloud-only migration checklist without implying local repository transfer	Coding and desktop	Framework backends
Business review workflow	Summarize an aging-report description before any committing outreach	Business workflows	Durable tasks
Scheduled or queue-driven work	Run one weekly report under a stable business idempotency key	Task automation	Mission lifecycle
Supervised multi-step agent	Create a resource-bounded research mission, signal it, and resume its events	Multi-step workflows	Mission lifecycle
Agent tool host	Register the 12 stable MCP wire contracts and keep approvals enabled	MCP integration	MCP tools
Existing-memory agent	Merge Mem0 facts, LangGraph documents, and graph edges without a migration	Memory federation	Memory federation
Operations dashboard	Reconnect SSE, correlate run state and usage, and inspect webhook deliveries	Monitoring and observability	Observability
Event-driven backend	Receive signed terminal run events without exposing an API key to a browser	Webhooks	Signed webhooks
Existing agent framework	Add governed tools or read-only memory without restarting the stack	Agent frameworks	Framework tutorials
Product capability discovery	Map a user outcome to a live product surface or documented developer contract	Capability Atlas	Product guides

Durable task automation

Use Execution Fabric when your application needs a narrow, deployed task contract with a durable run resource.

curl https://api.praxa.io/v1/execute \
  -H "Authorization: Bearer $PRAXA_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: weekly-review-2026-08-13" \
  -d '{
    "apiVersion": "v1",
    "requestId": "weekly-review-2026-08-13",
    "mode": "task",
    "task": { "input": "Summarize this week and propose the next safe action." },
    "idempotencyKey": "weekly-review-2026-08-13"
  }'

The 202 response confirms admission or exact idempotent replay. It is not proof that a downstream effect occurred. Continue with the returned run URL or event stream.

Governed mission orchestration

Use @praxa/sdk when your organization operates the versioned Integration Gateway and needs mission budgets, signals, resumable events, capability discovery, or governance readback.

import { PraxaClient } from "@praxa/sdk";

const praxa = new PraxaClient({
  baseUrl: process.env.PRAXA_BASE_URL!,
  accessToken: () => process.env.PRAXA_ACCESS_TOKEN!,
});

const mission = await praxa.createMission(
  {
    goalSpec: { task: "Prepare the weekly review" },
    resourceBudget: {
      maximumSteps: 12,
      maximumToolCalls: 8,
      maximumElapsedMs: 120_000,
      maximumParallelism: 2,
    },
  },
  "weekly-review-2026-08-13",
);

Existing-memory compatibility

Use @praxa/sdk/memory when memory already lives in Mem0, LangGraph, Zep, Graphiti, Letta, OpenAI Agents sessions, or another system. The federation engine is read-only and returns source-level success, timeout, unsupported, and failure states.

const result = await memory.recall({
  query: "Which editor does this developer prefer?",
  namespace: { tenantId: "acme", subjectId: "developer-123" },
  limit: 10,
});

if (result.status === "failed") {
  throw new Error("No memory source completed successfully");
}

An ok result means every selected source succeeded. partial means at least one source succeeded and at least one did not. failed is returned normally when none succeeded; provider failures do not become an empty successful result.

Continue with a tutorial

Run a durable task

Integrate the v1 task contract with cURL, JavaScript, Python, Next.js, or a Cloudflare Worker and verify it end to end.

Federate memory

Connect existing provider clients, preserve tenant boundaries, and test partial degradation and provenance.

Use MCP tools

Receive webhooks

Create a signed endpoint, preserve the raw request body, and test replay handling.

Integrate an agent framework

Choose Vercel AI SDK, OpenAI Agents, LangGraph, MCP, or a custom adapter without restarting your tool or memory stack.

Browse product capabilities

Explore all provided product outcomes with public statuses, surfaces, and testable examples.

Get started

Core Concepts

Releases

Praxa use cases

Choose by job

Apply the contracts to a product

Durable task automation

Governed mission orchestration

Existing-memory compatibility

Continue with a tutorial

Run a durable task

Federate memory

Use MCP tools

Receive webhooks

Integrate an agent framework

Browse product capabilities

​Choose by job

​Apply the contracts to a product

​Durable task automation

​Governed mission orchestration

​Existing-memory compatibility

​Continue with a tutorial

Run a durable task

Federate memory

Use MCP tools

Receive webhooks

Integrate an agent framework

Browse product capabilities

Choose by job

Apply the contracts to a product

Durable task automation

Governed mission orchestration

Existing-memory compatibility

Continue with a tutorial