# ElizaOS Weekly Newsletter (Apr 1–Apr 7, 2026) ## 1) Executive Summary This week was all about **shipping paths to production** and **tightening trust + reliability**—even if most of the work is still in review. - **Real-world deployment win:** Community member **jgonly1_89829** successfully deployed their first ElizaOS agent on **Hatcher.host** using a **one-click flow** with **Groq integration** and **live logs**—a strong signal that “agent hosting for normal humans” is getting easier. - **Message/action reliability improvements landed in a major PR (pending):** A large fix set addressing **TOON action parameters** and reducing **continuation-loop spam** was opened in `elizaos/eliza` (#6709). If merged, it should noticeably improve non-streaming connectors (Discord/Milady-style) and async task flows. - **Security + trust direction sharpened:** Two notable trust-oriented threads emerged: an issue proposing **delegation chains** for autonomous agents (scoped authority + spend limits), and a report that the **eliza MCP server** received an **A security grade** after passing all evaluated criteria. ## 2) Development Updates ### In review: TOON schema + async task loop fixes PR **#6709 “Fix/toon action params”** targets two pain points in `DefaultMessageService`: - **TOON action params now requested explicitly:** Adds a `params` field to the single-shot TOON response schema so the LLM actually returns required parameters (e.g., `RUN_IN_TERMINAL.command`) for connectors using TOON encapsulation. - **Stops noisy continuation loops for async task actions:** Marks actions like `CREATE_TASK`, `START_CODING_TASK`, `CODE_TASK`, `SPAWN_AGENT`, and `SPAWN_CODING_AGENT` as terminal in the continuation logic—preventing repeated filler responses while PTY sessions run in the background. This PR is big (template migration and utilities are included), but the user-facing impact is straightforward: **fewer “why didn’t the action get its required inputs?” failures** and **less connector spam** during long-running tasks. Resource: https://github.com/elizaos/eliza/pull/6709 ### In review: Windows checkout fix for plugin-openrouter A PR was opened in `elizaos-plugins/plugin-openrouter` to fix **PGlite memory artifacts blocking Windows Git checkouts**—a classic papercut that disproportionately affects new contributors on Windows. (No merge yet, but it’s a high-leverage quality-of-life fix.) ### In review: Local dev harness (“agent/” workspace) proposal PR **#6702** proposes a local REPL-style **agent harness** (`agent/`) plus a submodule workflow for local plugin development. Review notes flag install/workspace consistency issues (submodule workspaces committed unintentionally; lockfile mismatches). If those are ironed out, this could become the “fastest way to boot the repo and test a character” for contributors. Resource: https://github.com/elizaos/eliza/pull/6702 ### Proposed/active issues: Trust primitives for autonomous agents - **Delegation chains proposal:** A new issue suggests implementing **delegation chains** so agents can operate with *scoped authority*, *spend limits*, and *cascade revocation*—a concrete direction for safer autonomous behavior and composable trust. - **Security assessment:** The MCP server named **eliza** reportedly earned an **A grade** on Loaditout, passing **7/7 criteria**—useful validation as more integrations rely on MCP endpoints. ## 3) Community Spotlight (Discord) ### “I deployed my first agent” moment (Hatcher.host) A standout community highlight: **jgonly1_89829** shared that deploying on **Hatcher.host** was smooth via **one-click deployment**, with a clean dashboard and **live logs**, plus **Groq** wired in. For many community members, hosted deployment is the difference between “cool framework” and “usable product,” so posts like this are valuable field reports. Discussion link: https://discord.com/channels/1253563208833433701/1253563209462448241 ### AI News pipeline work: regen + fallbacks + credit monitoring **dankvr** shared ongoing work to regenerate about a week of AI news data, with two practical improvements planned: - **Fallback models** (for resilience when a primary model fails or rate limits) - **Better credit usage monitoring** (to prevent silent budget drain and improve observability) This is the kind of “boring reliability” work that keeps community-facing feeds stable. ### Debug corner: X (Twitter) login failing in eliza-agent In **#coders**, **dave_24736** reported an `eliza-agent` login failure on X: repeated retries leading to a “page not found” even with cookies. **odilitime** immediately asked for exact versions of **ElizaOS** and **plugin-twitter**, suggesting a potential version-specific regression or upstream change. If you’re seeing similar issues, add your version info when reporting—it’ll speed triage. Coders link: https://discord.com/channels/1253563208833433701/1300025221834739744 ### Project gardening: Elizok contributions **ipadkid.eth** asked whether **Elizok** is still active and offered to contribute; **odilitime** replied that it’s available on GitHub. If you’ve been waiting for a “starter contribution,” this is a good thread to pick up and turn into a concrete issue/PR list. ### Safety note: scammer warnings Moderation-wise, **airanna21** flagged multiple users as scammers during general discussion. Reminder: verify links, don’t share keys, and treat “urgent” DMs as hostile by default. ## 4) Token Economics (AI16z token + auto.fun) This week’s token conversation was dominated by **market sentiment** rather than protocol changes. In **#discussion**, community members debated the ELISA token’s sharp drawdown (claims of ~95% from highs), compared supply dynamics to **AI16z**, and argued about whether fundamentals can outrun macro conditions. What we *didn’t* see in the provided data: - No announced changes to **AI16z token** mechanics - No concrete shipping updates for **auto.fun** this week Takeaway: sentiment is noisy; **watch for real artifacts**—releases, revenue, usage, and governance proposals—rather than price-only narratives. ## 5) Coming Soon Based on what’s in motion right now, here’s what to anticipate: - **Improved connector compatibility via TOON params** (pending merge of #6709): fewer missing-parameter action failures for TOON-based connectors. - **Cleaner async task UX**: reduced continuation-loop spam during long-running orchestrations (also in #6709). - **AI news reliability upgrades**: fallback models + credit monitoring for the regeneration pipeline (work in progress). - **Trust model enhancements**: delegation chains proposal could evolve into a core primitive for safe autonomy (scoped authority, spend limits, revocation). - **Windows contributor experience fixes**: plugin-openrouter checkout patch should reduce friction for Windows devs once merged. ## 6) Resources **Discord** - General discussion (deployments, token talk, AI news): https://discord.com/channels/1253563208833433701/1253563209462448241 - Coders (X login issue): https://discord.com/channels/1253563208833433701/1300025221834739744 **GitHub (PRs / Issues)** - PR: TOON action params + async continuation-loop fixes (#6709): https://github.com/elizaos/eliza/pull/6709 - PR: Local dev “agent/” harness proposal (#6702): https://github.com/elizaos/eliza/pull/6702 - PR: MnemoPay “economic memory” plugin (in review, needs persistence + safety fixes) (#6701): https://github.com/elizaos/eliza/pull/6701 - Issue: Delegation chains for autonomous agents (scoped authority, spend limits, revocation): (see `elizaos/eliza` issues opened Apr 7) - Issue: MCP server “eliza” security grade A report: (see `elizaos/eliza` issues opened Apr 7) If you want a high-impact way to help this week: reproduce the **X login bug** with exact versions, or review **#6709** focusing on connector behavior and any edge cases introduced by the template migration.