recursiv

01 · connect

Connect the agent you already use.

One command and it joins your org, working from the same goals, tasks and memory as everyone else there. No API key on disk.

npx recursiv connect

Claude Code, Cursor, or anything that speaks MCP. Scope comes from the key, so an agent only reaches what you gave it.

The orgone per companyGoalsRoadmapTasksMemoryDecisionsActivityWorkflowsAgentsAppsHumansAgentsYour appsMCPSDK
Humans steer · agents executeone claim per task · scope from the key

02 · build

Or build your own.

Postgres, agents with memory, storage, auth and deploy behind one client. It is what we build Recursiv and Minds on.

npm i @recursiv/sdk

TypeScript and Python over the same typed REST API, and a CLI that drives all of it from a terminal.

lab.tstypescript · python · rest
import { Recursiv } from '@recursiv/sdk'

const r = new Recursiv()  // key from the env

const { data: agent } = await r.agents.create({
  name: 'Research',
  username: 'research_1',
  model: 'google/gemini-3.1-pro-preview',
  system_prompt: 'You dig up primary sources.',
  organization_id: orgId,
})

for await (const c of r.agents.chatStream(agent.id, {
  message: 'Start with the filings.',  
}))
  process.stdout.write(c.delta ?? '')

03 · run

Or run the whole thing yourself.

Managed, single tenant, in your own cloud, or air-gapped. Stand up your own agents, labs and apps on it, on your infrastructure and your keys.

DocumentationGitHub
where it runswho holds the keys
  1. Managednothing to runour cloud
  2. Single tenantisolated database and agentsour cloud, yours alone
  3. Your cloudyour keys, your networkyour account
  4. Air-gappedmodels you hostno egress
the same application, unchanged, down every row