@tinyzkp/cli

Generate and verify TinyZKP zero-knowledge proofs from your terminal. One command, zero setup.

npx @tinyzkp/cli templates
npx @tinyzkp/cli prove range_proof '{"min":0,"max":100,"witness_steps":[42,44]}' --wait > proof.json
npx @tinyzkp/cli verify proof.json

That's the whole flow.

Install

# Run without installing (recommended for one-off use)
npx @tinyzkp/cli <command>

# Or install globally
npm install -g @tinyzkp/cli
tinyzkp <command>

Requires Node.js 18+ (uses built-in fetch).

Authentication

Get a free API key at tinyzkp.com/signup — 100 proofs/month, no credit card. Then expose it via any of:

# Option 1: environment variable
export TINYZKP_API_KEY=tzk_xxxx

# Option 2: per-command flag
tinyzkp prove ... --api-key=tzk_xxxx

# Option 3: credentials file
mkdir -p ~/.tinyzkp
echo "TINYZKP_API_KEY=tzk_xxxx" > ~/.tinyzkp/credentials
chmod 600 ~/.tinyzkp/credentials

Commands

| Command | Description | |---|---| | tinyzkp templates | List all available proof templates | | tinyzkp describe <id> | Show parameters + example for a template | | tinyzkp estimate <id> <params> | Estimate cost / time / proof size (no key needed) | | tinyzkp prove <id> <params> | Submit a proof job, returns job_id | | tinyzkp poll <job-id> | Check status / fetch completed proof | | tinyzkp verify <proof> | Verify a proof from file or inline JSON | | tinyzkp healthz | Probe the API |

Common flags

| Flag | Purpose | |---|---| | --wait | On prove/poll: block until proof is complete | | --timeout=<s> | Max seconds to wait (default 300) | | --json | Machine-readable JSON output (good for piping) | | --zk | Enable zero-knowledge masking on prove | | --api-key=<key> | Override env var | | --base-url=<url> | Use a non-default API host (e.g., for staging) |

End-to-end example

# 1. List what's available
$ npx @tinyzkp/cli templates
Available proof templates:

  range_proof         (lightweight)
    Prove a secret value lies within a range
  hash_preimage       (lightweight)
    Prove knowledge of a hash preimage
  ...

# 2. See how to use one
$ npx @tinyzkp/cli describe range_proof
range_proof
backend: vm

Prove a secret value lies within a range

Parameters:
  min (integer) required
    Lower bound of the allowed range (inclusive)
  max (integer) required
    Upper bound of the allowed range (inclusive)
  witness_steps (array) required
    Additive steps from min that sum to (value - min)

Example:
  {"min":18,"max":120,"witness_steps":[7]}

# 3. Estimate before paying
$ npx @tinyzkp/cli estimate range_proof '{"min":0,"max":100,"witness_steps":[42,44]}'
Estimate
  trace length:     128 steps
  cost (Developer): $0.05
  proof size:       12 KB
  prove time:       1200 ms

# 4. Generate
$ npx @tinyzkp/cli prove range_proof '{"min":0,"max":100,"witness_steps":[42,44]}' --wait
✔ proof completed (prf_a1b2c3)
  version: 4
  size:    12.4 KB
  bytes:   0x6a8f7c4b1e9d2a3f5e8b6c9d2a3f5e8b6c9d2a3f5e8b...

# 5. Verify (always free)
$ npx @tinyzkp/cli prove ... --wait --json > proof.json
$ npx @tinyzkp/cli verify proof.json
✔ valid  (round-trip 8 ms)

Pipe-friendly mode

Add --json to any command for machine-readable output. Plays nicely with jq:

JOB=$(tinyzkp prove range_proof '{"min":0,"max":100,"witness_steps":[42]}' --json | jq -r .job_id)
tinyzkp poll "$JOB" --wait --json | jq .proof.bytes

What about MCP?

If you're building an AI agent (Claude Code, Claude Desktop, Cursor, OpenAI agents), prefer the MCP integration — your agent gets these same operations as native function calls without shelling out to a CLI:

claude mcp add --transport http tinyzkp https://mcp.tinyzkp.com

See tinyzkp.com/docs for the MCP integration guide.

License

MIT — see LICENSE.

Links

• tinyzkp.com — homepage
• tinyzkp.com/try — browser playground (no signup)
• tinyzkp.com/docs — full docs
• github.com/logannye/hc-stark — open source