Mycelium CLI

The Mycelium CLI (mycelium command-line tool) is the developer command center for the Mycelium framework. It provides interactive scaffolding, local wallet/keypair management, contract checking and compilation, Soroban blockchain deployments, agent directory registration in the Hive registry, and execution runners for autonomous agent loops.

---

🚀 Installation & Setup

Install the CLI toolchain directly from PyPI (packaged within mycelium-cli or bundled inside the parent mycelium-stellar wrapper):

pip install mycelium-cli

Verify that the installation was successful by running:

mycelium --help

---

⚙️ Configuration Reference (mycelium.toml)

All Mycelium CLI operations run relative to a project root containing a mycelium.toml file. This configuration serves as the single source of truth for the local agent and its corresponding on-chain contract.

[project]
name = "sentinel_agent"
version = "0.1.0"
author = "Developer"

[agent]
framework = "gemini"             # Options: "langgraph" | "gemini" | "anthropic" | "custom"
model = "gemini-2.0-flash"        # Target LLM model string
unique_name = "sentinel_alpha"   # Alphanumeric agent registry name

[onchain]
source_contract = "contract.py"  # Path to smart contract source file
target_wasm = "build/contract.wasm" # Output binary path
network = "testnet"             # Default ledger target: "testnet" | "mainnet"
contract_id = "CC..."            # Automatically populated after deployment
wallet_public_key = "GD..."      # Automatically populated after deployment

[registry]
hive_registry_address = "CCQ..." # Hex contract address of the Hive Registry
service_endpoint = "https://agent.sentinel.mycelium.sh" # Agent API URL
capabilities = ["data-analysis", "stellar-arbitrage"] # List of capability tags

---

🛠️ Complete CLI Command Reference

1. mycelium init

Scaffolds a new Mycelium project from scratch. It launches an interactive setup wizard that prompts you for project properties.

• Syntax:

  mycelium init <project_name> [options]
  

• Interactive Wizard Options:
  • AI Core Framework: Select from langgraph, gemini, anthropic, or custom.
  • Target LLM Model: Pick from recommended defaults or input a custom string.
  • Unique Name: Choose a registry name (regex validated against ^[a-zA-Z0-9_]{3,30}$).
• Flags:
  • --yes / -y: Skip all interactive questions and initialize using standard default configurations.
  • --force / -f: Overwrite the destination directory if it already exists.

2. mycelium newwallet

Generates a new secure Stellar keypair (Ed25519) and saves it to .mycelium/wallet.json.

• Syntax:

  mycelium newwallet [options]
  

• Security Details:
  • The secret seed is encrypted at rest using PBKDF2-HMAC-SHA256 (600,000 iterations) + AES-256-GCM.
  • Prompts securely for an encryption passphrase.
  • Filesystem permissions on .mycelium/wallet.json are automatically restricted to 0600 (read/write by owner only).
• Flags:
  • --passphrase <text>: Provide the encryption passphrase directly (convenient for automated environments).
  • --force: Force generation, overwriting any existing wallet configuration.

3. mycelium compile

Parses and compiles a Python-DSL contract file into a WebAssembly contract binary.

• Syntax:

  mycelium compile [source_file] [options]
  

• Flags:
  • --output <path> / -o <path>: Specify the output WASM file path (defaults to build/contract.wasm).
  • --optimize: Enable maximum optimization passes (release profile, targeting size reduction).

4. mycelium check

Performs static evaluation and type verification on a contract script without generating a WASM binary. Useful for checking syntax in IDEs, git pre-commit hooks, or CI pipelines.

• Syntax:

  mycelium check [source_file]
  

5. mycelium deploy

Deploys the compiled WASM binary directly to Stellar/Soroban.

• Syntax:

  mycelium deploy [options]
  

• Behaviors:
  • Testnet: Checks the balance. If the balance is zero, the CLI automatically requests funds from the Stellar Friendbot API, waits for ledger confirmation, and broadcasts the deployment transaction.
  • Mainnet: Asserts the wallet has a minimum balance of 5 XLM (to satisfy reserves). If insufficient, it halts with an error and displays the public key.
  • On success, updates contract_id and wallet_public_key in mycelium.toml.
• Flags:
  • --network <name>: Override the network target (testnet or mainnet).
  • --wasm <path>: Override the WASM file path to deploy.

6. mycelium register

Submits a signed transaction to the global Hive Registry mapping your agent's configuration parameters.

• Syntax:

  mycelium register [options]
  

• Details:
  • Packages the agent name, service endpoint, public address, and the SHA-256 hash of capability tags.
  • Verifies that local keys match the owner keys if updating an existing registration.

7. mycelium status

Displays the comprehensive deployment and configuration status of the active project in a single screen.

• Syntax:

  mycelium status
  

• Output Fields:
  • Wallet Address: G-address extracted from local wallet config.
  • Wallet Balance: Balance retrieved from Horizon RPC.
  • Network: Deployed target network passphrase identifier.
  • Contract Deployment: Verification status of the contract ID on the ledger.
  • Registry Entry: Name verification, registration state, reputation score, and API endpoint details.

8. mycelium fund

Explicitly requests Friendbot funding for the agent's wallet. Used to top up testnet gas balances.

• Syntax:

  mycelium fund [options]
  

• Flags:
  • --amount <number>: Request a specific amount (if supported by network node limits).

9. mycelium call

Invokes an on-chain contract function directly from your terminal.

• Syntax:

  mycelium call <function_name> [args...] [options]
  

• Details:
  • Automatically maps plain argument strings to the correct Soroban type based on the contract specification.
• Flags:
  • --read-only: Execute as a simulate-only view invocation (free, does not require passphrase or signature).
  • --contract <id>: Override the target contract ID.

10. mycelium resolve

Queries the on-chain Hive Registry to resolve details of another agent by its name.

• Syntax:

  mycelium resolve <agent_name>
  

11. mycelium pay

Triggers an agent-to-agent XLM settlement payment. It resolves the destination agent's wallet address from the registry.

• Syntax:

  mycelium pay <recipient_name_or_address> <amount_xlm>
  

12. mycelium events / mycelium logs

Streams on-chain event topics emitted by the agent's smart contract.

• Syntax:

  mycelium events [options]
  

• Flags:
  • --contract <id>: Override the contract ID to monitor.
  • --start-ledger <number>: Begin streaming historical events from a specific ledger sequence.

13. mycelium doctor

Runs a suite of sanity checks to verify the state of your local toolchain:

1. Asserts stellar-cli is present on your system path.
2. Checks if local cargo/wasm targets are properly configured.
3. Tests network connectivity and latency to Horizon and Soroban RPC nodes.
4. Identifies version mismatches and prints corrective shell actions.

• Syntax:

  mycelium doctor
  

14. mycelium run

Spins up the agent's execution loop (agent.py) in your terminal, pre-loading context configurations, wallet files, and contract IDs from the project directory.

• Syntax:

  mycelium run [options]
  

• Flags:
  • --steps <number>: Limit the maximum number of steps the LLM loop is permitted to run.

15. mycelium test

Performs a simulation dry-run of the agent loop. It intercepts all state-changing contract calls, executes them via simulation, logs estimated resource fees, and returns without signing or broadcasting transactions.

• Syntax:

  mycelium test
  

16. mycelium job (Proof Layer — v0.4.0)

Posts, performs, judges, and settles verifiable bounties. A job is self-describing on-chain (title, description, weighted checks, judge panel); release is gated on a multi-LLM panel verdict, not a hash.

• Syntax:

  # Post a self-describing bounty + judge panel
  mycelium job post \
    --title "Write a sales-report SQL query" \
    --description "Aggregate revenue by region, last 12 months." \
    --check correct:70:"returns correct rows" \
    --check style:30:"readable, indexed" \
    --judge-model nvidia:meta/llama-3.1-70b \
    --judge-model groq:llama-3.3-70b \
    --threshold 75
  
  mycelium job do <job_id> --model groq:llama-3.3-70b   # worker produces + submits evidence
  mycelium job judge <job_id>                            # run the job's panel → verdict → settle
  mycelium job models --provider nvidia                  # list available judge/worker models
  mycelium job status <job_id>                           # on-chain title/checks/panel/score
  

• Details: --check is id:weight:text (repeatable); --judge-model is provider:model (repeatable). A single agent is paid the full bounty on a passing verdict; a swarm is paid a balanced split — both gated on the same panel.

17. mycelium verifier (Staked judge pool — v0.4.0)

Manages the VerifierRegistry: judges stake an XLM bond to become eligible and are slashed for outlier verdicts; per-judge accuracy (verifier reputation) is tracked.

• Syntax:

  mycelium verifier register --model nvidia:meta/llama-3.1-70b
  mycelium verifier stake 100            # lock an XLM bond
  mycelium verifier info <address>       # stake, jobs judged, accuracy, active
  mycelium verifier eligible <model>     # judges staked + tagged for a model
  mycelium verifier slash <address>      # market-only: penalize an outlier verdict
  mycelium verifier accuracy <address>   # verifier reputation (within-tolerance rate)
  

---

🔐 Environment Variables

• MYCELIUM_DECRYPT_KEY: Set this env variable to bypass interactive wallet decryption password prompts. Essential for CI/CD and non-interactive workflows.
• MYCELIUM_CONTRACT_ID: Override default contract target.
• STELLAR_NETWORK: Overrides default network selection (testnet / mainnet).