Live
Memory and recall
Agent Brain
Self-hosted shared memory, semantic cache, and governance for Claude, Codex, Cursor, Gemini, Aider, and custom agents.
Build with agentswithout losing the plot.
Four connected products for real AI-assisted development work: capture notes by voice, persist project memory, route tasks across agents, and operate from a purpose-built terminal. Studio and Avatar are internal projects in development used to produce the marketing material for the suite.
What this site is selling
Not vague AI optimism. A suite built around real developer pain: lost context, fragmented tools, weak verification, and terminals that were never designed for assisted implementation.
agent-software://suiteworkflow stack
wispr.capture("ship the harness rollout notes")
brain.remember(decision)
harness.assign(workerTask)
terminal.run(verification)
// fewer broken handoffs, fewer lost decisions4
live product surfaces, with Studio and Avatar in development
1
sign-in across the suite experience
Local-first
local models, local memory, and local project context stay at the center of the workflow
Online access
account login and suite access require an internet connection
Release status
Agent Brain, Agent Wispr, Agent Harness, and Agent Terminal are live. Agent Studio and Agent Avatar are internal projects in development used to produce the suite's marketing material, not available products or services. The suite is local-first in how it works, but account access and sign-in still depend on an internet connection.
See it in action
Watch the suite work, not just read about it.
Agent Terminal overview
See voice capture, project memory, and bounded task delegation working together in one session.
Products
Start with the bottleneck that is slowing you down.
Some teams need shared memory first. Others need faster note capture, cleaner multi-agent execution, or a terminal that can keep up with how they actually work. The suite is strongest together, but each product is meant to solve a separate, expensive failure mode.
Buy the product that removes today's friction. Add the others when the workflow around it becomes the next constraint.
DeepSeek Harness integration
Agent Harness gives DSH a verified execution loop.
DeepSeek Harness provides the local, composable runtime. Agent Harness sits above it as the planning, routing, approval, and verification layer. The free dsh-agent-harness plugin connects the two so bounded coding work can use your own local models without giving up evidence or operator control.
Run it offline with your DSH configuration, choose local or cloud providers per task, and keep the same verification contract as the rest of the suite.
agent-software://dsh-bridgefree · MIT · offline
dsh plugin --profile local add \ github:agentsoftwaresuite/dsh-agent-harness#v0.1.0 harness route task \ --worker dsh --verify # local provider, shared evidence
How it fits
Four apps, one development loop.
Wispr captures the thought, Brain supplies the context, Harness performs bounded and verified work, and Terminal keeps the human in the loop across the whole handoff. The result is a workflow where intent, memory, execution, and evidence can move together instead of being rebuilt at every tool boundary.
1
Agent Wispr
Capture intent without breaking flow
Speak a bug report, implementation idea, decision, or task handoff. Wispr turns it into usable text locally so the thought can move straight into the development loop instead of getting lost in a scratchpad.
2
Agent Brain
Recall the context that makes the request useful
Brain searches durable project memory, semantic cache entries, and learned execution patterns, then gives Terminal and connected agents the relevant context without stuffing every past session into the prompt.
3
Agent Harness
Run bounded work with evidence
Harness routes a focused task to a worker, keeps the scope and expected result explicit, applies policy, runs verification, and reports the run status instead of treating a worker's confidence as proof.
4
Agent Terminal
Keep the operator in control
Terminal is the bridge: it receives the composer draft, asks Brain for reusable answers and rewrite suggestions, runs the pre-send guard, launches bounded Harness work, tracks the result, and writes useful execution outcomes back to Brain.
Operating posture
Local-first where it matters. Honest about where it does not.
This is not a fake privacy pitch. The suite now requires account access, but the working core still prioritizes local models, local context, and self-hosted control over the parts that shape day-to-day development.
Local by default, not cloud-first by accident
Brain, Harness, and Terminal are designed around running close to your machine and your repo, so the default workflow is not "send everything to somebody else's dashboard."
Internet for access, local models for the work
The suite now requires sign-in, so it is not a fully offline product. The actual coding workflow still centers on local models, local storage, and your own development environment.
Project memory stays under your control
Agent Brain stores long-lived context in infrastructure you control, so recall, retention, and deletion are decisions you make instead of defaults inherited from a hosted assistant.
One account across the suite
Sign in once and move across Brain, Terminal, and Harness without re-authenticating on every hop. Credentials live in the OS keychain rather than in random local files.
Pricing
Buy the missing piece or buy the full operating system.
Start with one product
Pick the product that fits
Best when the pain is specific: Brain for context loss, Wispr for capture, Harness for agent coordination, Terminal for daily execution flow.
This is the right choice when one bottleneck is obviously more expensive than the others.
All-Access Pass
Every product, one subscription
Best when the work already crosses product boundaries every day: speak notes, preserve decisions, dispatch work, verify the result, and keep the whole loop inside one connected stack.
If your team is already stitching these capabilities together manually, the suite price is usually cheaper than the drag.
Updates
Shipping notes, positioning changes, and product moves.
2 min read
Two months of shipping the Agent Software Suite story: what we learned and what changes next
The most useful lesson from the eight-week run was that technical trust comes from clarity and honest limits more than from louder claims.
By Agent Software
2 min read
The future of AI coding infrastructure: memory, evals, terminals, and orchestration
The future of AI coding looks less like one perfect agent and more like a stack of infrastructure layers that make agents usable.
By Agent Software
Affiliate Program
Promote Agent Software Suite. Earn 30% recurring.
Refer paying builders and earn 30% on every invoice they pay for the next 12 months.
30%
Commission
12 invoices
Window
60 days
Cookie
Follow the suite as products ship.
Join Discord, browse product docs, or follow the feed for release notes and readiness updates.
Is there a suite bundle checkout?
Yes. The All-Access Pass covers Agent Brain, Agent Wispr, Agent Terminal, and Agent Harness under one subscription. Individual product plans are also available — see current pricing on the Pricing page.
Which products are live now?
Agent Brain, Agent Wispr, Agent Harness, and Agent Terminal are live today. Agent Studio and Agent Avatar are internal projects in development used to produce the suite's marketing material; neither is currently available or purchasable.
Where should product documentation live?
This site is the hub, not the dumping ground. It should point you to the right product docs quickly, while deeper implementation details stay on the product-specific sites.
How do I follow updates?
Use the blog when you want the reasoning, Discord when you want the fast signal, and the RSS or JSON feeds when you want to track shipping changes without checking the site manually.