A company's single AI platform: providers and models, employee
agents, tools, and spend — all governed in one place. Answers are
grounded in company knowledge, cited to their source and scoped to
each user's access rights.
Pillar 1 · Hub
All of a company's AI, in one place
A small or mid-size company that wants to give employees LLMs and
agents usually has to glue several tools together: a gateway for API
keys, a separate chat, a separate knowledge base, a separate agent
builder — and then write the glue between them. Achilles replaces
that zoo with a single system: connect your providers, configure
access, and everything just works.
On your own
A zoo of tools
Key gateway, chat, knowledge, agents — four systems
Between them, homegrown glue you have to maintain
Spend is smeared across tools — no unified picture
Access and permissions differ in every tool
Onboarding an employee means four accounts
With Achilles
Connect it and it works
Providers and models in a single registry
Spend is visible by employee and model; limits keep the budget
Every company agent in one catalog, under admin control
Tools are configured once — for both chat and agents
Web interface, Slack bot, and API out of the box
Providers and sources are connected once — AI is then available to
employees on every surface
Providers and models
Connect by API key, activate models from the catalog, and assign
them by function: chat, agents, embeddings.
Spend under control
Token accounting by employee and model, weekly agent limits, and
an alert when the monthly budget is exceeded.
Employee agents
Employees build agents in the interface, no code required: task,
model, schedule. The admin sees them all and can pause any one.
Tools, centralized
Web search, page fetching, knowledge-base search. Availability for
chat and agents is configured in one place; connecting MCP servers
is coming soon.
AI behavior
The organization's system prompt — the company's tone, rules, and
context — is edited right in the admin panel.
Surfaces
Web interface out of the box, a Slack bot, MCP for external AI
assistants, and a Public API for your own surfaces — Telegram and
other messengers get added as you grow.
Access and security
Roles, email invitations, API keys, an audit log — user management
is built in, not bolted on the side.
Agent library — soon
Preset agents for common tasks: company analytics, digests,
monitoring — turn one on and use it.
The hub is useful from day one — employees get models and agents
without a single line of code. But the real depth appears once the
platform knows your company.
Pillar 2 · Knowledge
Answers grounded in company knowledge
A company's context is scattered across Jira, Slack, Confluence, and
drives — for every answer, everyone goes to everyone, and the bigger
the team, the more this coordination costs. Achilles connects to your
sources and builds a single knowledge graph: people, projects, tasks,
and decisions linked in one place.
Example: how the graph links people, projects, tasks, and documents
C = n(n − 1) ÷ 2
6 people = 15 links
→
C = n
6 people = 6 links
500 people: 124,750 everyone-to-everyone links
→ 500 queries to the platform
One query instead of ten conversations
Instead of quizzing colleagues and combing through Jira,
Confluence, and Slack — one question to the platform, with an
answer linked to the source.
Access rights are respected
The platform carries source permissions into the graph: an
employee's answer only contains what they have access to.
Knowledge stays with the company
The history of decisions is preserved in the graph — even after
key people leave.
Onboarding in days, not months
A new hire gets the full picture from day one — without dozens of
intro meetings.
Example
How it works in practice
A new developer wants to understand why the notifications service was
built on PostgreSQL rather than MongoDB.
Without Achilles
4 days, 0 answers
Search Confluence — nothing
Ask in Slack — "don't remember, ask Alex"
Alex is on vacation
Decision lost, time wasted
With Achilles
1 query, full context
Discussion in #backend (March 2024)
Related ticket INFRA-891
ADR: write load, ACID requirement
Decision author: @alex, DBA: @john
Rollout
The platform grows with your knowledge base
Value doesn't require "connect everything first." Achilles is useful
from day one as an AI hub and matures into a grounded company memory
as the knowledge base fills up — with no mode switches and no
reinstall.
Base empty
AI hub
Deploy, add people and messengers, pick a model. Chat and agents
answer from general knowledge — out of the box, with no
integrations and no code.
→
Base filling up
Hybrid
First sources connected. A question about company matters gets a
grounded answer with a link to the source; everything else works
like a consultant.
→
Base full
Company memory
A full knowledge graph under unified access. Every answer is
grounded in corporate data, and agents work across the company's
entire context.
The transition is seamless because there's no mode to configure: with
an empty base, the model simply isn't given the search tool, and the
platform answers like an assistant; the first ingested chunk turns
grounding on by itself. No flag, no switch — the barrier to entry
drops from weeks to an hour.
Open Source
Open code, full control
Achilles is an open-source platform: every line can be read and
verified. Deployment is flexible — in the cloud or inside your
company's own perimeter.
Auditable code
No black boxes. Security you can verify.
Data under control
When needed, the entire platform runs inside your perimeter — data
never leaves the company.
Summary
One platform instead of ten tools
Achilles is a single infrastructure for AI in a company: a hub of
models, agents, and spend that matures into company memory as sources
are connected.
AI governance hub
Providers, models, tools, and spend — in a single registry.
Company knowledge
A knowledge graph from your sources, with answers that respect
access rights.
Agents without code
Employees create, the platform runs, the admin governs.
Access wherever it's convenient
Web interface, Slack, MCP, Public API — out of the box.