Give an agent access

This is the golden path: you create a secret, register an agent, grant it read access via a policy, then the agent fetches the secret at runtime.

Try it out

Try out the example in this repo: Basic (SDK vault + secret + policy) and LangChain Agent (agent fetches secrets just-in-time).

1. Create a vault and secret (human)

Log in (email/password or Google), then:

curl

# Get token (see Quickstart for humans)
export TOKEN="..."

# Create vault
VAULT_RESP=$(curl -s -X POST https://api.1claw.xyz/v1/vaults \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name":"Production","description":"Prod secrets"}')
VAULT_ID=$(echo "$VAULT_RESP" | jq -r '.id')

# Store a secret
curl -s -X PUT "https://api.1claw.xyz/v1/vaults/$VAULT_ID/secrets/api-keys/openai" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"type":"api_key","value":"sk-proj-..."}'

TypeScript

import { createClient } from "@1claw/sdk";

const client = createClient({
  baseUrl: "https://api.1claw.xyz",
  apiKey: process.env.ONECLAW_API_KEY,
});

const { data: vault } = await client.vault.create({
  name: "Production",
  description: "Prod secrets",
});

await client.secrets.set(vault.id, "api-keys/openai", "sk-proj-...", {
  type: "api_key",
});

2. Register an agent (human)

curl

AGENT_RESP=$(curl -s -X POST https://api.1claw.xyz/v1/agents \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name":"My Bot","description":"CI agent","scopes":["vaults:read"]}')
AGENT_ID=$(echo "$AGENT_RESP" | jq -r '.agent.id')
API_KEY=$(echo "$AGENT_RESP" | jq -r '.api_key')
# Store API_KEY securely; it is shown only once.

TypeScript

const { data } = await client.agents.create({
  name: "My Bot",
  description: "CI agent",
  scopes: ["vaults:read"],
});
const agentId = data.agent.id;
const apiKey = data.api_key; // Store securely — shown only once

3. Create a policy (human)

Grant the agent read access to all secrets in the vault (or use a narrower pattern like api-keys/*):

curl

curl -s -X POST "https://api.1claw.xyz/v1/vaults/$VAULT_ID/policies" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d "{
    \"secret_path_pattern\": \"**\",
    \"principal_type\": \"agent\",
    \"principal_id\": \"$AGENT_ID\",
    \"permissions\": [\"read\"]
  }"

TypeScript

await client.access.grantAgent(vault.id, agentId, ["read"], {
  secretPathPattern: "**",
});

4. Agent fetches the secret

From the agent's environment (with AGENT_ID and API_KEY stored securely):

curl

# Get agent JWT
AGENT_TOKEN=$(curl -s -X POST https://api.1claw.xyz/v1/auth/agent-token \
  -H "Content-Type: application/json" \
  -d "{\"agent_id\":\"$AGENT_ID\",\"api_key\":\"$API_KEY\"}" | jq -r '.access_token')

# Fetch secret
curl -s "https://api.1claw.xyz/v1/vaults/$VAULT_ID/secrets/api-keys/openai" \
  -H "Authorization: Bearer $AGENT_TOKEN"

TypeScript

import { createClient } from "@1claw/sdk";

// In the agent's runtime
const agentClient = createClient({
  baseUrl: "https://api.1claw.xyz",
  agentId: process.env.ONECLAW_AGENT_ID,
  apiKey: process.env.ONECLAW_AGENT_API_KEY,
});

const { data: secret } = await agentClient.secrets.get(VAULT_ID, "api-keys/openai");
// Use secret.value — don't log or persist

The response includes the decrypted value. The agent uses it for the intended call and does not persist or log it.

Summary

Step	Who	Action
1	Human	Create vault, store secret
2	Human	Register agent, save API key
3	Human	Create policy: agent + path pattern + read
4	Agent	Get token, GET secret by path

To revoke: delete the policy or deactivate the agent. To rotate: create a new secret version (PUT) or rotate the agent key.

Alternative: Agent self-enrollment

Instead of steps 1-2 above, the agent can self-enroll by calling a public endpoint with no credentials. With human_email, Allow/Deny links are emailed (and the response may include approval_url as a backup). With name only, the response includes approval_url for you to open while signed in:

curl -s -X POST https://api.1claw.xyz/v1/agents/enroll \
  -H "Content-Type: application/json" \
  -d '{"name":"my-agent","human_email":"you@example.com"}'

Or name only:

curl -s -X POST https://api.1claw.xyz/v1/agents/enroll \
  -H "Content-Type: application/json" \
  -d '{"name":"my-agent"}'

Or via the CLI:

npx @1claw/cli agent enroll my-agent --email you@example.com
npx @1claw/cli agent enroll my-agent

After you approve the pending enrollment, you receive the agent's credentials by email and still need to create a policy (step 3). After that, the agent proceeds with step 4 as normal.

This is particularly useful when:

• The agent is deployed independently and doesn't have access to the human's dashboard.
• You're onboarding many agents that each need to register with their human counterpart.
• The agent is an AI assistant that wants to store and share secrets with its user.

See Agent Self-Onboarding for the full agent-first flow, including sharing secrets back to the human.

For managing large numbers of agents, see Managing Agent Fleets.