Solid

Solid is an autonomous agent platform that gives AI agents their own devices (Windows, macOS, Linux, iPhone, Android), accounts (Google, Apple), and spending budgets to complete complex, multi-step jobs from start to finish. Users describe a goal in plain language, set approval rules, and close their laptop — agents handle the rest and message back when the work is done. The platform is available as a web product and via API for teams that want to embed agents inside their own products.

What It Is

Solid sits in the autonomous AI agent category, distinct from simple chatbots or single-step automation tools. Rather than requiring prebuilt connectors or step-by-step orchestration, Solid's agents determine which tools a job needs, handle setup and authentication, build missing integrations, and self-correct when something breaks. The result is delivered as a finished artifact — a hosted link, a CRM update, a pull request — rather than a list of instructions for the user to execute.

How the Autonomy Spectrum Works

Solid describes three layers of agent capability that distinguish it from simpler automation:

• Self-sufficient — Agents choose tools, sign up for services, and pay within approved budgets. They can operate any tool a person can use: APIs, websites, desktop software, or phone apps.
• Self-healing — When a tool fails or a setup breaks, agents investigate, repair, and verify the job can continue. If they remain blocked, they explain what needs human input.
• Self-improving — Agents retain lessons from corrections and apply them to future jobs, building a shared context across the team's work.
• Self-scaling — Agents can spin up additional Solid agents or coordinate external agents (such as Codex or Claude Code), divide work, and return one checked result.

Example Workflows

The homepage illustrates five concrete use cases that show the range of jobs Solid's agents can handle:

• Custom sales demos — Convert customer meeting notes into a hosted, tested demo app and return the link.
• LinkedIn lead generation — Research accounts against an ideal customer profile, draft outreach for approval, follow up, book meetings, and update the CRM.
• AI product evaluations — Build and run evals after every release, simulate users completing key tasks, judge outcomes, and send a report.
• Production bug resolution — Investigate an alert, reproduce the issue, write and test a fix, open a pull request for engineer approval, and verify recovery.
• Customer support resolution — Trace a stalled ticket across systems, apply a policy-compliant fix, and confirm resolution with the customer.

Deployment Model and Control

Solid offers three deployment options: Solid Cloud, a private VPC, and on-premises (availability depends on setup). Enterprise customers can configure access controls, spending budgets, approval workflows, and policies across a workspace. The API tier lets teams ship always-on agents inside their own products, with persistent context and long-running job support under developer-defined controls.

Agents are described as "meta-agents" — they can see and manage their own tools, devices, and spending, so users can ask at any time what a job cost, what is still running, and whether spending could be reduced.

Why It Matters

The platform's core proposition is that agents do not require prebuilt connectors or user expertise in the tools involved. According to the Solid website, if a person can use a tool, a Solid agent can too — whether that means calling an API, operating a website, running desktop software, or using a phone app. This positions Solid against connector-list automation platforms and against agent frameworks that require developers to wire up each integration manually.