Choose theme
AI Resources
Paperclip
Paperclip is a self-hosted server and dashboard for coordinating AI agents through assigned tasks, company goals, budgets, and approvals.
Agents bring their own models and runtimes. Paperclip tracks their work and scheduled runs through adapters for tools such as Claude Code, Codex, and HTTP agents. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A task system around existing agents
Issues connect assigned work to a company, project, goal, and parent task, with comments, documents, and attachments kept alongside it.
Why it stands out
Work and spending in the same dashboard
Operators can follow scheduled runs, review pending decisions, and inspect recorded costs by company, agent, or project.
Availability
Open-source and self-hosted
The current installation guide requires Node.js 24.11 or newer. A common local setup uses embedded PostgreSQL; agent runtimes and authentication need their own setup.
Why it matters
What makes it useful
When several agents contribute to one project, a task needs an owner and a place to leave partial progress. Paperclip's issue guide describes checking out work before starting, recording progress in comments, and identifying blockers so another person or agent can pick up the thread.
What to know
Where it fits
Its Decisions view is for work waiting on a person. The publisher describes proposals with a question and defined options, such as creating a task or changing its status. Dismissing a proposal records that response without applying its effects; it is different from selecting an action. Moving a company to another instance involves a separate check. The publisher's export guide says approvals, cost events, and activity logs are not included in company packages. Its Not included panel shows those counts before download, so someone needing that history can see what the package will leave behind before using it for a transfer.
Notable points
What stands out
The cost guide separates monthly company and agent budgets from lifetime project spending caps. The current README qualifies enforcement: it uses recorded spend, and reporting delays or work already running can delay a stop. A dashboard budget is therefore not a guarantee that a provider invoice will stay below that amount.
Before using
What to review
The Codex adapter guide says a managed agent home inherits the host's Codex login by default; a configured per-agent API key takes precedence. That choice changes which account the agent uses.
The repository publishes security advisories. An April advisory reports inherited Gmail-connector access through a Paperclip-managed Codex runtime for version 2026.403.0, and a July advisory reports a local-trusted deployment attack for versions below 0.3.1. Neither lists a patched version; these reports alone do not establish the status of a current installation.
The v2026.1001.0 release changes native and legacy agent execution to full-auto defaults when explicit restrictive settings are absent. Paperclip's controller approvals are separate from runtime tool permissions. The current Codex guide scopes its approval and sandbox bypass to the classic CLI engine; the ACP path keeps its writable-workspace sandbox with network access by default. Review the actual engine and permission configuration before scheduling work.
Check the server's retained task records, attachments, runtime credentials and outputs, who can access them, and your backup and deletion arrangements. Local installation data persists after the command exits. The publisher distinguishes trusted-local loopback use from authenticated network deployments; choose the authentication and network boundary for the intended users.
The README says first-party usage telemetry is enabled by default and documents opt-out settings, including PAPERCLIP_TELEMETRY_DISABLED=1 or DO_NOT_TRACK=1. Its description excludes issue content, prompts and secrets from those events; that is the publisher's stated boundary. Separately configured OpenTelemetry trace export is an opt-in observability path.
Review what scheduled heartbeats and adapter tools can do with the supplied accounts. Company decisions, spend records and a successful connection probe do not by themselves restrict every file, command or provider action.
Company export packages are not full-instance backups. The export guide's history omissions need a separate retention plan; the README also warns that package exports can retain plain environment values and local paths even when referenced secrets are omitted.
Reader fit
Who may find it relevant
People coordinating several agents on continuing projects may find the task, schedule, cost and decision records useful. Operating it requires configuring the server and each agent runtime. Someone who only needs one interactive assistant has less need for company and scheduled-task administration. Builders connecting a Codex runtime can use the adapter's Test Environment check for the command, working directory, authentication, and a hello probe before scheduling runs. This checks the connection; a successful probe does not show that an assigned task has been completed.
Editorial note
Why LifeHubber lists it
We include Paperclip for people coordinating continuing agent work because it makes recovery from a stranded run part of the tracked task. Recovery actions preserve the original assignee and record the evidence, recovery owner, wake policy and resolution; unsafe or exhausted retries are escalated to the board. That gives operators a place to follow what happened to interrupted work instead of losing that explanation in a silent reassignment.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Agent-Reach
Panniantong/Agent-Reach
A CLI routing layer that helps command-capable agents reach web, social, repository, video, RSS, and search sources through ordered backends. For multi-backend channels, its doctor command reports the active backend. The current main guide describes check-only default setup; the v1.5.0 guide uses --safe for checking without installing core infrastructure.
Genex Desktop
genex-games/genex-desktop
An early desktop workspace for making browser games with AI, combining game chat, playable previews, build history, assets, configurable model roles, local Blender tools, and static web export.
PageIndex
VectifyAI/PageIndex
A vectorless, reasoning-based RAG framework for long-document retrieval, tree-structured indexing, traceable document search, and agent context workflows.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
