AgentBoards briefing · refreshed daily

pi: Added an oauth.authServerMetadataUrl setting for MCP servers that advertise a wrong OAuth authorization server or none.

Check the release's compatibility and migration changes before upgrading.

Latest lead story: Oct 1, 2026 · Listed agents only. Updates, newcomers and practical comparisons.

The short version

pi — Added an oauth.authServerMetadataUrl setting for MCP servers that advertise a wrong OAuth authorization server or none. Pi uses the configured metadata document instead of discovery.
Automagik Genie — New saved workflow test-simplify audits a test suite and proposes an ordered campaign of batches to delete or consolidate low-value tests. It is read-only and proposal-only; one refuter per owner group tries to keep each candidate.
Paperclip — Migrations 0280–0283 run on upgrade (agent personas, routine webhooks, Slack communication guidance, GitHub review bots).
OpenAI Agents SDK — mcp: add configurable MCP listing page limits
Agent Orchestrator (AO) — Add mobile review workflow
Dify — Open-source platform for building agentic workflows, RAG pipelines and LLM apps on one canvas

Major & notable updates · past 30 days

In focus · Major update· TerminalOct 1, 2026

pi: Added an oauth.authServerMetadataUrl setting for MCP servers that advertise a wrong OAuth authorization server or none.

Earendil's minimal, MIT-licensed terminal coding agent and toolkit, extended with TypeScript extensions and skills instead of MCP

What changed. Added an oauth.authServerMetadataUrl setting for MCP servers that advertise a wrong OAuth authorization server or none. Pi uses the configured metadata document instead of discovery.

Added quietStartup: "header", which keeps the startup header with version and key hints but hides the model scope line and loaded-resource listing.

Added a copy code login method to Anthropic /login for headless setups where the browser runs on another machine

Why it matters. Check the release's compatibility and migration changes before upgrading.

Where it stands out

Supports local models. Supports headless or CI workflows. Lets you bring your own model.

Before you switch

Free and MIT-licensed; sign in to subscription providers with /login or supply your own API keys

Runs local · macos, linux, windows.

Compare with Codex CLI and Claude Code

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	pi	Codex CLI	Claude Code
Runs	local	local	local / cloud
Pricing	Free and MIT-licensed; sign in to subscription providers with /login or supply your own API keys	Included with ChatGPT Free, Go ($8), Plus ($20), Pro (from $100), Business and Enterprise; or pay per token with an OpenAI API key	Included with Claude Pro and Max subscriptions, or pay-as-you-go through an API key
Local models	Supported	Supported	Not listed
Open source	MIT	Apache-2.0	No

pi vs Codex CLI →pi vs Claude Code →

Explore pi →Read the source Story link

Notable update· HarnessOct 2, 2026

Automagik Genie: New saved workflow test-simplify audits a test suite and proposes an ordered campaign of batches to delete or…

Interviews you into a plan, dispatches coding agents in parallel, then reviews the result against acceptance criteria

What changed. New saved workflow test-simplify audits a test suite and proposes an ordered campaign of batches to delete or consolidate low-value tests. It is read-only and proposal-only; one refuter per owner group tries to keep each candidate.

New ## Owner approvals section in AGENTS.md: a change the owner was not asked about goes in its own question, and an approved decision is reopened only by a question that quotes it and shows old → new.

Why it matters. Check whether the new capability fits a task you already use this agent for.

Where it stands out

Supports multi-agent workflows. Lets you bring your own model. Open source · MIT.

Before you switch

Free and open source under MIT; it dispatches the coding agents and subscriptions you already run

Runs local · macos, linux.

Upstream calls it Genie; it is listed here as Automagik Genie to keep it distinct from Cosine Genie, an unrelated product already on the board.

Genie plans and dispatches; the model is whatever the coding agents it hands work to are configured with.

Compare with Trinity and GitLab Duo

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	Automagik Genie	Trinity	GitLab Duo
Runs	local	local / cloud / sandbox	local / cloud
Pricing	Free and open source under MIT; it dispatches the coding agents and subscriptions you already run	Free and open source under Apache-2.0; you bring an Anthropic or Google API key for the agents it runs	Duo Core is included with Premium and Ultimate subscriptions; Duo Pro and Duo Enterprise are per-seat add-ons, and Agent Platform usage is billed with GitLab Credits
Local models	Not listed	Not listed	Supported
Open source	MIT	Apache-2.0	No

Automagik Genie vs Trinity →Automagik Genie vs GitLab Duo →

Explore Automagik Genie →Read the source Story link

Notable update· HarnessOct 2, 2026

Paperclip: Migrations 0280–0283 run on upgrade (agent personas, routine webhooks, Slack communication guidance, GitHub review…

Open-source control plane that organises Claude Code, Codex, OpenClaw, Hermes and other agents into a company with budgets and approvals

What changed. Migrations 0280–0283 run on upgrade (agent personas, routine webhooks, Slack communication guidance, GitHub review bots).

Why it matters. Check whether the new capability fits a task you already use this agent for.

Where it stands out

Supports multi-agent workflows. Open source · MIT.

Before you switch

Free and MIT-licensed to self-host on a laptop, in Docker or on a server; agents use your own subscriptions or keys; cloud deployments are listed as in progress on the roadmap

Runs local / cloud · macos, linux, web.

Compare with Trinity and GitLab Duo

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	Paperclip	Trinity	GitLab Duo
Runs	local / cloud	local / cloud / sandbox	local / cloud
Pricing	Free and MIT-licensed to self-host on a laptop, in Docker or on a server; agents use your own subscriptions or keys; cloud deployments are listed as in progress on the roadmap	Free and open source under Apache-2.0; you bring an Anthropic or Google API key for the agents it runs	Duo Core is included with Premium and Ultimate subscriptions; Duo Pro and Duo Enterprise are per-seat add-ons, and Agent Platform usage is billed with GitLab Credits
Local models	Not listed	Not listed	Supported
Open source	MIT	Apache-2.0	No

Paperclip vs Trinity →Paperclip vs GitLab Duo →

Explore Paperclip →Read the source Story link

Notable update· FrameworkOct 2, 2026

OpenAI Agents SDK: mcp: add configurable MCP listing page limits

OpenAI's lightweight Python framework for multi-agent workflows with handoffs, guardrails, sessions, tracing and MCP

What changed. mcp: add configurable MCP listing page limits

sandbox: add opt-in Docker removal protection

migrate Vercel sandboxes to the owned provider client

Why it matters. Worth comparing if you connect your agent to external tools and services.

Where it stands out

Supports local models. Supports headless or CI workflows. Supports multi-agent workflows.

Before you switch

Free and MIT-licensed; you pay OpenAI API usage, plus per-call fees for hosted tools such as web search, file search and code interpreter

Runs local / cloud · macos, linux, windows.

Compare with Google Agent Development Kit (ADK) and Claude Agent SDK

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	OpenAI Agents SDK	Google Agent Development Kit (ADK)	Claude Agent SDK
Runs	local / cloud	local / cloud	local / cloud
Pricing	Free and MIT-licensed; you pay OpenAI API usage, plus per-call fees for hosted tools such as web search, file search and code interpreter	Free and Apache-2.0 licensed; you pay for Gemini or another model provider and any Google Cloud deployment resources	Free SDK; you pay Anthropic API usage (or Bedrock, Vertex AI, Foundry); claude.ai subscriptions may not be used for third-party products
Local models	Supported	Supported	Not listed
Open source	MIT	Apache-2.0	MIT (SDK); use governed by Anthropic Commercial Terms

OpenAI Agents SDK vs Google Agent Development Kit (ADK) →OpenAI Agents SDK vs Claude Agent SDK →

Explore OpenAI Agents SDK →Read the source Story link

Notable update· HarnessOct 1, 2026

Agent Orchestrator (AO): Add mobile review workflow

Open-source desktop IDE from Untrivial AI that plans, delegates and supervises fleets of coding agents on a live kanban board

What changed. Add mobile review workflow

Why it matters. Check whether the new capability fits a task you already use this agent for.

Where it stands out

Supports multi-agent workflows. Open source · Apache-2.0.

Before you switch

Free and Apache-2.0 licensed desktop app; agents use your own subscriptions or keys; no paid plan published

Runs local · macos, windows, linux.

Compare with Orca and BitFun

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	Agent Orchestrator (AO)	Orca	BitFun
Runs	local	local	local
Pricing	Free and Apache-2.0 licensed desktop app; agents use your own subscriptions or keys; no paid plan published	Free and MIT-licensed; agents run on your own Claude, OpenAI or other subscriptions and keys	The core code is MIT and the app is free to download; you configure a provider API key and pay that provider
Local models	Not listed	Not listed	Not listed
Open source	Apache-2.0	MIT	MIT

Agent Orchestrator (AO) vs Orca →Agent Orchestrator (AO) vs BitFun →

Explore Agent Orchestrator (AO) →Read the source Story link

New to AgentBoards · past 30 days

Recently added to the directory. A listing date is not a product launch date.

New to the board· FrameworkSep 4, 2026

Dify joins AgentBoards

Open-source platform for building agentic workflows, RAG pipelines and LLM apps on one canvas

Why it matters. New to this directory—not necessarily newly launched. Compare its workflow and requirements with tools in the same category.

Where it stands out

Supports local models. Supports multi-agent workflows. Lets you bring your own model.

Before you switch

Community edition free to self-host; Dify Cloud Sandbox free, Professional $590/year, Team $1,590/year, Enterprise on request

Runs local / cloud · macos, linux, windows, web.

Compare with n8n and Langflow

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	Dify	n8n	Langflow
Runs	local / cloud	local / cloud	local / cloud
Pricing	Community edition free to self-host; Dify Cloud Sandbox free, Professional $590/year, Team $1,590/year, Enterprise on request	Self-hosted community edition free; cloud Starter EUR 20/month, Pro EUR 50/month, Business EUR 667/month billed annually, Enterprise custom	Open source and free to self-host, with a free hosted cloud account available
Local models	Supported	Supported	Supported
Open source	Dify Open Source License (Apache-2.0 with additional conditions)	No	MIT

Dify vs n8n →Dify vs Langflow →

Explore Dify →Official website Story link

New to the board· HarnessSep 4, 2026

Coder joins AgentBoards

Self-hosted platform for running coding agents and cloud development environments on your own infrastructure

Why it matters. New to this directory—not necessarily newly launched. Compare its workflow and requirements with tools in the same category.

Where it stands out

Supports local models. The whole platform is self-hosted, so an on-premises or in-VPC model endpoint can be used through the AI Governance gateway. . Includes Docker sandboxing. Supports headless or CI workflows.

Before you switch

Community edition is free and open source and self-hosted; Premium is priced annually per user and Enterprise is custom

Runs cloud / sandbox · macos, linux, windows, web.

Coder runs whichever agent and model the team chooses; the AI Governance gateway brokers the LLM calls rather than supplying a model.

The whole platform is self-hosted, so an on-premises or in-VPC model endpoint can be used through the AI Governance gateway.

Compare with Trinity and Ellipsis

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	Coder	Trinity	Ellipsis
Runs	cloud / sandbox	local / cloud / sandbox	cloud / sandbox
Pricing	Community edition is free and open source and self-hosted; Premium is priced annually per user and Enterprise is custom	Free and open source under Apache-2.0; you bring an Anthropic or Google API key for the agents it runs	Usage-based with no per-seat fee: model tokens at cost plus a 10% platform fee, $0.142/vCPU-hour and $0.024/GB-hour; optional support packages from $5,000/month; open-source free tier
Local models	Supported	Not listed	Not listed
Open source	AGPL-3.0	Apache-2.0	No

Coder vs Trinity →Coder vs Ellipsis →

Explore Coder →Official website Story link

New to the board· TerminalSep 4, 2026

Zero joins AgentBoards

Go terminal coding agent with 25+ providers, permission and sandbox policy, MCP client and server, and a scriptable exec mode

Why it matters. New to this directory—not necessarily newly launched. Compare its workflow and requirements with tools in the same category.

Where it stands out

Supports local models. Supports headless or CI workflows. `zero exec` returns exit codes and has a documented GitHub Action recipe. . Supports multi-agent workflows.

Before you switch

Free and open source under MIT; bring your own provider key or run a local model

Runs local / sandbox · macos, linux, windows.

Zero ships a native Linux sandbox helper using seccomp rather than Docker; source builds must build the `zero-linux-sandbox` helper separately.

Browser and terminal control helpers ship with the npm and release builds; source builds need those helper binaries on PATH or configured in local-control settings.

Compare with Factory Droid and Gemini CLI

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	Zero	Factory Droid	Gemini CLI
Runs	local / sandbox	local / cloud / sandbox	local / sandbox
Pricing	Free and open source under MIT; bring your own provider key or run a local model	Pro $20/mo, Plus $100/mo, Max $200/mo, Teams $60/mo plus $40 per seat; Business and Enterprise custom; usage governed by rolling rate limits	Paid Gemini API key (per-token), Vertex AI, or Gemini Code Assist Standard/Enterprise (1,500-2,000 requests/day); free and Google AI Pro/Ultra access discontinued June 18, 2026
Local models	Supported	Supported	Supported
Open source	MIT	No	Apache-2.0

Zero vs Factory Droid →Zero vs Gemini CLI →

Explore Zero →Official website Story link

New to the board· IDESep 4, 2026

Theia IDE joins AgentBoards

Eclipse Foundation's open, vendor-neutral AI-native IDE, built on the Theia AI framework and any LLM you point it at

Why it matters. New to this directory—not necessarily newly launched. Compare its workflow and requirements with tools in the same category.

Where it stands out

Supports local models. Ollama, llamafile and any OpenAI-compatible endpoint such as vLLM are documented LLM providers. . Supports multi-agent workflows. Theia ships several named chat agents plus an Orchestrator agent that routes a request to the right one. . Lets you bring your own model.

Before you switch

The IDE and framework are free and open source; you pay only for the LLM provider you configure

Runs local / cloud · macos, linux, windows, web.

Ollama, llamafile and any OpenAI-compatible endpoint such as vLLM are documented LLM providers.

Theia ships several named chat agents plus an Orchestrator agent that routes a request to the right one.

Compare with Kiro and Cursor

Same-category alternatives, matched by execution, platform and capabilities. Open the full comparison for source dates and qualifications.

Profile	Theia IDE	Kiro	Cursor
Runs	local / cloud	local / cloud / sandbox	local / cloud
Pricing	The IDE and framework are free and open source; you pay only for the LLM provider you configure	Free tier with 50 credits/mo; Pro $20 (1,000 credits), Pro+ $40 (2,000), Pro Max $100 (5,000), Power $200 (10,000) per user/mo; add-on credits $0.04 each	Free Hobby tier; Pro $20/mo, Pro+ and Ultra with 3x/20x Pro limits; Teams $40/user/mo; Cloud Agents billed at API pricing; custom API keys supported for chat models
Local models	Supported	Not listed	Not listed
Open source	EPL-2.0	No	No

Theia IDE vs Kiro →Theia IDE vs Cursor →

Explore Theia IDE →Official website Story link

Release log · other recent releases

Version tracking, separate from featured stories.

Cline 3.0.68Oct 2, 2026
Sudo Code v0.2.21Oct 2, 2026
graff v0.0.302.14Oct 2, 2026
Pydantic AI 2.53.0Oct 2, 2026
CopilotKit 4.23.1Oct 2, 2026
Hermes Studio 0.7.27Oct 2, 2026
Cua cua-spaces-v0.2.0Oct 2, 2026
Scream Code 0.17.7Oct 2, 2026
Pane v2.4.149Oct 2, 2026
Oh My Pi v18.4.10Oct 2, 2026
eve 0.70.0Oct 2, 2026
oh-my-openagent 5.1.9Oct 2, 2026
SLICC 6.234.0Oct 2, 2026
Google Agent Development Kit (ADK) 2.11.0Oct 2, 2026
Code Puppy 0.0.890Oct 2, 2026
Ruflo v3.50.0Oct 2, 2026
Omnara 1.0.21Oct 1, 2026
minion 3.23.0Oct 1, 2026
Sortie v1.26.0Oct 1, 2026
WrongStack v1.0.30Oct 1, 2026