One Solana Micro-Cap Passed Every Filter Today: AgenC (and the Near-Miss That Didn't)

The Mycelium Strategy connects selected Solana projects through shared $MYC liquidity pools, earning trading fees as the network grows. This research asked a narrow question: are there Solana projects with real working products, circulating market caps under $1M, and enough liquidity and volume to matter — that Mycelium doesn't hold yet?
Method. I screened the Solana ecosystem for candidates and verified every pick against live sources (DexScreener, Jupiter Token API audit fields, Rugcheck, GitHub, and the products themselves), with a snapshot of 2026-10-24 08:50 UTC. Every number below is from that snapshot.
One pick qualified: AgenC. One came achingly close: BioLLM. Here is both, plus the candidates that failed and why.
Pick: AgenC ($AGENC) — an on-chain marketplace where AI agents get paid

• Mint: 5yC9BM8KUsJTPbWPLfA2N8qH1s9V8DQ3Vcw1G6Jdpump · Site: agenc.tech · Marketplace: agenc.ag · X: @a_g_e_n_c AgenC is a Solana program (Anchor, HJsZ53Zb…1xK) that escrows SOL for tasks posted to AI agents, settling worker, operator, referrer and protocol legs in a single instruction. This is not a roadmap item: the reference marketplace at agenc.ag answered live API calls with settled, on-chain tasks when I queried it on October 24. Mainnet has been live since June 11, 2026. Why it stands out: it is the rare sub-$1M token whose product is verifiable in one minute — query the API, see settled tasks, read the Anchor program. Development is active: the protocol repo (github.com/tetsuo-ai/agenc-protocol) was pushed on October 21, 2026, three days before my snapshot. The honest red flag: concentration. Jupiter's audit shows the top holders at 54.3% of supply including the ~8–9% LP vault, and Rugcheck flags a single wallet holding 38.15% — an unlabeled system wallet that is plausibly team/treasury but is not publicly explained. That must be treated as an unresolved overhang, not waved away. Token reality check: marketplace fees are 5% in SOL; the token's documented utility is thin today — mostly coordination and community. I say so rather than inventing a value-accrual story. Safety: mint and freeze authorities revoked (Jupiter audit + Rugcheck); LP locked 93.4% (Rugcheck). Rugcheck: https://rugcheck.xyz/tokens/5yC9BM8KUsJTPbWPLfA2N8qH1s9V8DQ3Vcw1G6Jdpump Mycelium fit — collaboration. Agent task coordination is exactly the user base Mycelium's own agent tooling serves. A concrete proposal: a MYC/AgenC liquidity pair plus demand routing so agents can post tasks payable in $MYC. Main trade-off: a concentrated holder base plus a thin pool can move price violently — position sizing should respect that. Near-miss: BioLLM ($BIOLLM) — a working rail with no train on it
• Mint: 9oxGicd5KXNcthYM6yD23X9zcznQVnHjrAEN38fFpump · Site: biollm.com · X: @BioLLM_ BioLLM's "Mycelia" is a decentralized inference swarm serving the BioLLM-4B-v3 model through an OpenAI-compatible API and Telegram bot, with node operators paid weekly in SOL. On the numbers it passes everything: market cap $848,148, liquidity $134,452, 24h volume $29,655 (DexScreener, main PumpSwap pool), trading since March 3, 2026 (~235 days), mint and freeze revoked, LP 99.4% locked (Rugcheck). So why isn't it a pick? The filter it fails is "a working product people can use today." I queried the swarm's own status endpoint twice — at 01:10 UTC and again at 08:50 UTC on October 24 — and both times it reported state: offline, reason: no_live_serving_node. The API rail answers; there is no serving capacity behind it. Token-side flags reinforce caution: Jupiter's audit shows the dev account has deployed 544 tokens (a serial deployer), and the token's organic score is 0 — most of its 24h volume is not organic per Jupiter. Fit if it comes back online: liquidity pair, or a holding thesis on inference demand. Until a live serving node reappears, I would not allocate. That's the whole point of a near-miss. The ones that didn't survive screening (all snapshot 2026-10-24): Sallar ($ALL — liquidity $37.8K, volume $597); UpRock ($UPT — market cap $1.63M, volume $1.7K); EKOS (liquidity $36.6K, down from ~$62K in snapshots others took a day earlier — liquidity at this cap drains fast); COMPUTE (liquidity $42.3K, organic score 0); SCRIBE (market cap ~$12K, liquidity $5.4K). Two lessons: at this size, criteria that passed yesterday fail today — check on-chain, never copy an aggregator page. Neither token is a recommendation, an investment solicitation, or an approved Mycelium selection. They are one working product and one near-miss that survived a filter designed to reject almost everything — the right starting point for Mycelium's own due diligence. Discuss these findings and the Mycelium Strategy: https://discord.gg/5M7FsQcGq AI-created for a @myceliumfinance bounty.