Paperclip
Paperclip proposes persistent agent chats that hand plans into ordinary project tasks
A four-part Agent Chat stack would give each person an ongoing conversation with an agent while retaining task controls and durable project handoffs. The experiment stays off by default and is distinct from the team-oriented Conference Room.
Paperclip is working on a middle ground between a disposable assistant chat and a task that must be completed after every reply. Its new Agent Chat stack uses ordinary company-visible tasks as the backing store for persistent person-to-agent conversations. The conversation can clarify a request and develop a plan before handing bounded work to a normal assigned project task.
The facts
- Backend PR #13282 proposes one company-visible conversation task per person and agent, gated by enableAgentChat. - Idle conversations stay out of normal task lists, recovery work and execution rollups rather than being treated as unfinished jobs. - Ordered /new resets create session boundaries without deleting retained history or waking idle conversations. - Handoff plans are persisted before assigned tasks are scheduled, and project/repository actions use the responsible person's existing grants. - Runtime PR #13281 carries the conversation directive through native, ACP, legacy and gateway paths, including OpenClaw and Hermes requests. - UI PR #13283 reuses task composers, transcripts and side panels and adds project-created cards, session dividers and gated sidebar shortcuts.
Why it matters
Planning conversations and execution tasks have different completion semantics. A friendly check-in should not become a stranded task that a recovery loop wakes overnight. Conversely, an accepted plan should not remain only in chat prose when another agent is expected to execute it. Reusing the task infrastructure keeps permissions, artifacts and history together, but it makes reset ordering and late provider writes especially important. The roadmap provides the product direction; the stack shows how that direction is being divided into reviewable layers.
Current
Inspected on 2026-09-12. The Paperclip release baseline is v2026.831.1 published 2026-09-02T04:56:26Z. The main source was open (not merged) when captured. Proposals, merged mainline changes and packaged releases are different availability states; the linked release baseline does not imply that open proposals have shipped.
Evidence
Primary evidence: paperclipai/paperclip PR #13282 (https://github.com/paperclipai/paperclip/pull/13282). Supporting context: paperclipai/paperclip PR #13281 (https://github.com/paperclipai/paperclip/pull/13281); paperclipai/paperclip PR #13283 (https://github.com/paperclipai/paperclip/pull/13283); Paperclip — CEO Chat product roadmap (https://raw.githubusercontent.com/paperclipai/paperclip/master/ROADMAP.md). Sources were read directly through public pages or GitHub API bodies. Contributor test results are attributed claims, not independently reproduced results.
Operator take
Review this as one dependent feature rather than four independent additions. The documented order is runtime, backend, UI and end-to-end coverage. An operator test should clarify and revise a request, create a project handoff, stop a reply, reset the session and confirm that old context does not leak into the new conversation. The contributor's paid campaign covered an earlier implementation across four adapter lanes; the source explicitly does not claim that campaign was repeated on every rebased head. Local and CI reports supply additional evidence, not permission to erase that version boundary.
Caveat
The inspected stack was proposed, not a generally available stable feature. Agent Chat is disabled by default and is separate from Conference Room. Company-visible backing tasks are not a promise of private one-to-one chat, and the database and queue changes need integrated qualification before a production rollout.
The inspected stack was proposed, not a generally available stable feature. Agent Chat is disabled by default and is separate from Conference Room. Company-visible backing tasks are not a promise of private one-to-one chat, and the database and queue changes need integrated qualification before a production rollout.