Evaluate a subnet before staking

Goal: decide whether staking into a subnet right now is a reasonable idea — not just "is it healthy," but "who already controls it, and what would my stake actually buy."

The sequence

1. get_subnet { netuid }              — identity, integration readiness, curation state
2. get_subnet_health { netuid }       — is it actually up right now
3. get_subnet_economics { netuid }    — price, pool size, emission share, registration status
4. get_subnet_concentration { netuid } — who already holds the stake (Nakamoto coefficient)
5. get_subnet_stake_quote { netuid, amount, direction: "stake" } — what your amount actually buys

Steps 1–4 are read-only lookups; run them in any order once you have a netuid. Step 5 needs a real amount, so run it last.

What each output means

• get_subnet — profile.readiness.score (0–100) and curation.review_state are the fastest signal for "is this a real, maintained subnet" before you look at anything financial.
• get_subnet_health — summary.status should be ok. A subnet that's degraded or failed right now doesn't disqualify it from staking (health tracks the subnet's API surfaces, not its chain economics), but it's worth knowing before you go looking for its dashboard.
• get_subnet_economics — registration_allowed + open_slots tell you if new miners/validators can even join; alpha_price_tao and emission_share are the subnet's current market size.
• get_subnet_concentration — the number that actually answers "how risky is this stake." stake.nakamoto_coefficient is how many entities you'd need to collude to control the subnet. 1 means one entity already can.
• get_subnet_stake_quote — price_impact_pct tells you whether your own stake would move the price meaningfully; expected_out is the alpha you'd actually receive.

Failure branch

get_subnet_health returning unknown for every surface (not ok/degraded/failed) means the prober hasn't reached this subnet's endpoints yet — check get_subnet's curation.review_state instead; a maintainer-reviewed subnet with no recent probe data is usually just new to the registry, not actually down.

Executed transcript (subnet 1, "Apex", 2026-07-28)

// 1. get_subnet { netuid: 1 }
{ "name": "Apex", "status": "active",
  "profile": { "readiness": { "score": 96, "readiness_tier": "buildable" },
               "curation_level": "maintainer-reviewed" } }

// 2. get_subnet_health { netuid: 1 }
{ "summary": { "status": "ok", "ok_count": 3, "failed_count": 0, "avg_latency_ms": 218 } }

// 3. get_subnet_economics { netuid: 1 }
{ "alpha_price_tao": 0.008532047, "registration_allowed": true, "open_slots": 0,
  "validator_count": 10, "miner_count": 246, "emission_share": 0.006692 }

// 4. get_subnet_concentration { netuid: 1 }
{ "stake": { "nakamoto_coefficient": 1, "gini": 0.833289, "top_1pct_share": 0.585467 } }

// 5. get_subnet_stake_quote { netuid: 1, amount: 100, direction: "stake" }
{ "expected_out": 11725.4, "expected_out_unit": "alpha", "price_impact_pct": 0.384 }

Reading this one: readiness (96) and health (ok) both look good, and 100 TAO would only move the price 0.38% — but nakamoto_coefficient: 1 means a single entity already holds enough stake to control consensus on this subnet. That's not disqualifying (many active subnets look like this today), but it's the one number this sequence surfaces that a glance at the price chart wouldn't.

Tools used

get_subnet, get_subnet_health, get_subnet_economics, get_subnet_concentration, get_subnet_stake_quote — same tools, callable directly over MCP (prompts/get name="evaluate_subnet_before_staking") or as the REST routes each links to.