v0.1 — MVP · Harness-as-a-Service context layer

The context layer
for your code.

Prompts are assembled, not written. Context is pulled, not remembered.

Take the free assessment See the framework →
Phase 1 — Prompt library
·
Model-agnostic · BYOK
·
Open core · GitHub
·
No tracking · No cookies
// The problem

Speed doesn't have to
mean drift.

Every decision made in a chat thread, every assumption in someone's head, every update that never touches documentation — that's context debt accumulating in real time.

01

Tribal Knowledge Failure

Domain context lives in one person's head and never gets captured. When that person is heads-down, the team stalls.

02

Documentation Drift

Governing docs are written once and go stale the moment work begins. Nobody trusts them, so nobody reads them.

03

Bi-directional Blindness

Code evolves without informing docs. Docs change without reaching developers. The two never meet.

// How it works

dovetell prompts.
you assemble.

Four inputs. One assembled context. Delivered into your flow. The loop closes when you review what the system captures.

01
Personal (local)
Your prompts live in your repo. Coding guidelines, style preferences, personal context.
available now
02
Team (shared)
Team prompts versioned with the team. Architecture, security, standards.
available now
03
Conversation context
Decisions in progress pulled from your team's tools — GitHub, Slack, Linear.
coming soon
04
Context base
Living markdown files. Decisions, assumptions, definitions. Unvetted until you approve.
coming soon
See the full framework diagram →
// Where does your team stand?

Four levels.
One clear path forward.

Take the free Team AI Maturity Assessment. 5 minutes. Get your score, your gaps, and exactly what to do next.

Scattered
0 – 22
Ad hoc
AI use is isolated. Context lives in individual chats. No shared standards.
Structured
23 – 54
Repeatable
Some practices exist but they're inconsistent. Handoffs vary. Knowledge drifts.
Coordinated
55 – 72
Aligned
Shared context, standard prompts, consistent handoffs. dovetell's target entry point.
Compounding
73 – 90
Systematic
AI context compounds over time. Every decision becomes searchable precedent.
Find out where your team scores. 5 minutes. 6 capability areas. Personalized gaps and next steps.
Take the free assessment →
// Where dovetell lives

The context layer in
your agent harness.

dovetell is not a model. It's not a harness. It's the context injection layer that makes your harness useful for your specific team's domain knowledge.

🧠
Model
Claude · GPT-4o · Gemini · Local (Ollama)
↓
Agent Harness
🛠
Tools + Skills
Cursor · Claude Code · GitHub Copilot · VS Code
🕊
dovetell — context injection layer
Personal prompts + team prompts + context base · model-agnostic · BYOK
💾
Memory + State
filesystem · AGENTS.md · git · context base
↓
</>
Your Codebase
The right context, right in your flow.
"if you're not the model, you're the harness" — LangChain, 2026  ·  "context graphs are AI's next trillion-dollar layer" — Foundation Capital, 2025
// The thinking behind it

Three frameworks.
One independent conclusion.

dovetell didn't start with a thesis. It started with a problem. These frameworks arrived later and named what was already being built.

LangChain · March 2026
"Harnesses today are largely delivery mechanisms for good context engineering. The question changed from what should we tell the model to what environment should the model operate in."
dovetell is the context injection layer in your agent harness. Phase 1 is manual. Phase 2 is automated.
Foundation Capital · Dec 2025
"The missing layer is decision traces — the exceptions, overrides, and precedents that currently live in Slack threads, deal desk conversations, and people's heads."
dovetell's context base is a decision trace store. "unvetted → reviewed → becomes truth" is the compounding loop.
AI Daily Brief · April 2026
"Harness-as-a-Service gives you three things: which model, which tools, which task. The domain context your team needs is not included."
Cursor handles the runtime. dovetell handles what the runtime needs to know about your team.
// Early access

Tell it once.
Let it travel.

Join the waitlist for the dovetell platform. Be the first to know when the context base, unvetted queue, and MCP integration go live.

✓ You're on the list. We'll be in touch when the platform launches.

No spam. No frequency. Just a note when it's ready.