In development - not yet available

One desktop client for every model you use

Forager is a multi-model desktop client for AI work. Chat, code, put models in a room together and let agents work unattended, with local models as a first-class option and your organization's rules enforced on the device.

Ask about early access See the app

Forager chat view: a conversation list on the left and an answer with a table and a YAML snippet on the right

A look at the app

Captured from the running development build. Conversations shown are sample data.

Roundtable dialog with two participants, Analyst and Skeptic, answering the same question — **Roundtable.** Several models in one conversation, each with its own persona, and a choice of whether they see each other's answers.

Command palette listing actions such as new conversation, compare models and new roundtable — **Command palette.** Every action and conversation is a keystroke away, including model comparison and roundtables.

Governance settings with allowed and blocked models and providers, an update channel and central feature flags — **Governance.** Allow or block models and providers, set an update channel and receive central feature flags from your organization.

Chat view with rendered markdown, a table and a code block — **Rich answers.** Markdown, tables, code, charts and diagrams render inline. Conversations are searchable, can be grouped into projects and can be exported.

What it does today

The feature set in the current development build.

Many providers, one client

Anthropic, OpenAI and Google Vertex, plus local models through Ollama and llama.cpp for work that should not leave the machine. An OpenAI-compatible gateway reaches AWS Bedrock and other brokered models.

Roundtable and compare

Convene several models, each with its own model and persona, plus an optional moderator. Share context so they build on each other, or isolate them and compare answers side by side.

A coding agent you review

Describe a change and the agent plans it with workspace-scoped tools, then stages the edits as a diff for your approval. Nothing is written or run without consent.

Isolated language servers

Editor diagnostics for TypeScript, Python, Rust, Go and ESLint. The Go, Rust and ESLint servers run in network-off containers pinned by digest, so the host needs no extra toolchains.

Dev in Docker

Optionally run the coding agent's commands inside a container with the workspace mounted, isolating the filesystem and network from your machine.

Tools, skills and knowledge

Connect MCP servers, load skills, index your own documents for retrieval, and keep snippets and presets for repeated work.

AI-powered artifacts

Generated documents and small apps can call a model through a mediated bridge that is consented per artifact and capped by budget.

Unattended runs

An autonomy mode with governance gates, checkpoints and recovery, so agents can keep working through a queue when you step away.

Quick ask and channels

Summon a quick-ask window with a global shortcut, publish self-contained deliverables locally, and reach an agent from a Telegram bot.

What's different

Most AI clients are a window onto one vendor. Forager is built around a few choices that change how it behaves.

No single vendor

Switching providers is a setting, not a migration. Local models are first-class, and a conversation that names an unavailable model falls back to one you have, with a notice.

Disagreement on purpose

A roundtable or a side-by-side comparison makes a second opinion routine. Isolated mode keeps the answers independent so you can see where models actually differ.

Consent before action

Agent edits arrive as a staged diff, tools are scoped to the workspace, and container isolation is available for both language servers and commands.

Standalone by default

Forager works fully on its own. Connecting it to a Hivemind tenant adds identity, policy and sync, and if the tenant is unreachable it falls back to standalone rather than locking you out.

Built for organizations

Controls an administrator sets centrally and the client enforces locally.

SSO (OIDC) and SCIM provisioning
Google sign-in
Model and provider allow and block lists
Central feature flags
Per-conversation and per-artifact budgets
Audit trail
Over-the-air updates verified by SHA-512 before install
Update channel control
Self-service account deletion with a grace period
Automated accessibility checks in CI

Interested in Forager?

Forager is in development and not yet available. Tell us what you would use it for and we will follow up as early access opens.

support@thehivemind.io