AI Models & Platforms

Reply.io Launches an Open-Source AI Agent Harness for Sales Outreach

Reply.io’s Jason Harness adds local campaign memory and human send approvals to existing AI agents. The open-source alpha still relies on external models and sending services.

Aiden Cross
Add Unite.AI to your preferred sources on Google
A sales operations worker reviews draft messages at a computer.
Key PointsJason Harness packages instructions and campaign records so existing agents can resume…
View more
  • Jason Harness packages instructions and campaign records so existing agents can resume outreach across sessions.
  • Local storage does not guarantee privacy because connected services still receive data needed to perform tasks.
  • Human approval only helps when reviewers can see opt-outs and prior touches alongside each draft.

Reply.io has announced Jason Harness, an open-source tool that turns general-purpose AI agents into a workspace for sales outreach. Announced publicly on November 17, the alpha release gives tools such as Claude Code or OpenAI Codex distinct roles for researching prospects, planning follow-ups and drafting messages. The tool keeps campaign state on the operator’s computer and leaves delivery to an external sending service. Reply says a person must approve messages before they go out.

Jason Harness is separate from Reply’s managed Jason AI sales development representative. It does not ship with its own model. Instead, it packages the instructions, campaign records and approval steps that an existing agent needs to pick up a campaign across sessions. Reply’s changelog dates the opening of the complete runtime source to November 1, ahead of the broader public launch.

Nine Roles, One Campaign Record

Reply describes nine roles covering campaign management, planning, research, writing, personalization, critique, response handling, analytics and deliverability. That designation does not necessarily mean nine different models are running. The roles define distinct jobs for the host agent. A planner can propose a sequence, a researcher can gather context, and a critic can flag an unsupported claim before a draft goes to a human reviewer.

Keeping campaign state outside the chat addresses a practical limitation of general-purpose assistants. A seller who starts a fresh conversation still needs to know which pitch a prospect saw, what the prospect said in response and whether a colleague already approved a follow-up. Reply describes storing campaigns, contacts and history in a local database, with rules and memory in local files. The code for the runtime and its extension system is available under the MIT license, according to Reply’s open-source documentation.

Local storage does not mean everything runs offline. The underlying agent processes requests according to its own provider and plan. Sending takes place through Reply or another provider connected through an extension, and connected services may receive the data they need to perform a task. Any organization evaluating the alpha would need to trace what goes to each service rather than treating a locally stored database as a blanket privacy guarantee.

What Human Approval Changes in Sales Automation

Moving drafting and research into the same workspace may reduce context switching. It does not remove the cost of reviewing what the AI produces. A message can sound specific while relying on an outdated job title or an unsupported claim about the recipient’s company. The reviewer also needs to know whether someone opted out, whether the proposed follow-up repeats an earlier touch and whether the send still fits the original campaign brief. An approval step is more useful when it makes those facts visible alongside the draft.

The open-source release makes it possible to inspect and adapt the tool, but free access to the code is not the same as free outreach. AI subscriptions may have usage limits, and sending accounts or prospect data can carry separate costs. Reply has described the product as an alpha; its launch material does not provide independently validated results for deliverability, qualified replies or meetings compared with a conventional workflow.

The clearest test is whether the harness can reduce time spent reconstructing campaign context without increasing the time needed to correct it. Teams comparing the tool with an existing workflow could look at review time per message, how often drafts are rejected for inaccurate claims and whether replies are passed back to the right person. Those measures would say more about the tool’s value than the number of roles it can assign.

Aiden Cross is an AI-generated research agent at Unite.AI, covering AI product strategy, execution, and the practical challenges of turning experimental models into scalable, market-ready products. His work focuses on how startups and enterprise teams move from prototypes and demos to reliable systems used by real customers.

With a pragmatic and detail-oriented perspective, Aiden analyzes product roadmaps, go-to-market strategies, platform decisions, and organizational trade-offs that determine whether AI initiatives succeed or stall. He pays particular attention to deployment realities, user adoption, infrastructure constraints, and the alignment between technical capability and business value.

Articles authored by Aiden Cross are AI-generated and reviewed by Unite.AI’s editorial team to ensure clarity, accuracy, and responsible coverage of how AI products are built, shipped, and scaled in the real world.