diff --git a/docs/architecture/README.md b/docs/architecture/README.md
new file mode 100644
index 000000000..9d4e4dd0f
--- /dev/null
+++ b/docs/architecture/README.md
@@ -0,0 +1,30 @@
+# Architecture Decision Records (ADR)
+
+This is a location to record all high-level architecture decisions in the Tendermint-RS project.
+
+You can read more about the ADR concept in this [blog post](https://product.reverb.com/documenting-architecture-decisions-the-reverb-way-a3563bb24bd0#.78xhdix6t).
+
+An ADR should provide:
+
+- Context on the relevant goals and the current state
+- Proposed changes to achieve the goals
+- Summary of pros and cons
+- References
+- Changelog
+
+Note the distinction between an ADR and a spec. The ADR provides the context, intuition, reasoning, and
+justification for a change in architecture, or for the architecture of something
+new. The spec is much more compressed and streamlined summary of everything as
+it is or should be.
+
+If recorded decisions turned out to be lacking, convene a discussion, record the new decisions here, and then modify the code to match.
+
+Note the context/background should be written in the present tense.
+
+To suggest an ADR, please make use of the [ADR template](./adr-template.md) provided.
+
+## Table of Contents
+
+| ADR \# | Description | Status |
+| ------ | ----------- | ------ |
+| [001](./adr-001-repo.md) | Repository structure for `tendermint-rs` | Proposed |
diff --git a/docs/architecture/adr-001-repo.md b/docs/architecture/adr-001-repo.md
new file mode 100644
index 000000000..9554d832e
--- /dev/null
+++ b/docs/architecture/adr-001-repo.md
@@ -0,0 +1,128 @@
+# ADR 001: Repository Structure
+
+## Context
+
+This crate began its life in the
+[tendermint/kms](http://github.com/tendermint/kms) 
+as a library of components necessary to build the Tendermint KMS. 
+Other features were added to support additional tooling, like a Tendermint config parser and RPC client. 
+
+We would now like to develop it further to support lite clients and eventually
+full nodes. It has thus been split into its own repo to evolve independently of
+the KMS. Here we propose an initial plan for the evolution of the repo towards
+these goals.
+
+## Decision
+
+At this stage, the repository must address three concerns:
+
+1) Maintain existing tendermint-rs roughly as is for the KMS and rpc client
+users
+2) Establish a Lite Client
+3) Establish a Reactor Framework and initial implementations of Tendermint
+reactor components
+
+For now, we can consider that each of these will proceed as their own crate.
+`tendermint-rs` already exists as the repo namesake, while `lite` and `reactor`
+crates can be created in the root.
+
+
+### Maintaining Tendermint-RS
+
+We should support the current tendermint-rs as necessary, and not hasten to make
+any sweeping changes to the structure until we better understand consumers.
+Any changes must be well co-ordinated with at least the KMS. Also, the
+`secret_connection` code should be 
+[moved back to the KMS](https://github.com/interchainio/tendermint-rs/pull/21#issuecomment-529061992)
+
+We may ultimately want to consider further breaking it up into crates. At
+present, that might consist of the following:
+
+- crypto
+- tendermint-rs (depends on crypto) 
+- config (depends on tendermint-rs)
+- rpc (depends on tendermint-rs)
+
+See some prior discussion about this
+[here](https://github.com/interchainio/tendermint-rs/issues/7).
+
+### Lite Client
+
+Most of the necessary data types already exist (pubkey/signature, blockid,
+vote), with support for serialization/deserialization. The crux of the lite
+client is verifying validator sets by computing their merkle root, and verifying
+commits by checking validator signatures. We have recently completed first
+passes at these verifications - they need to be further reviewed, better
+structured, and better tested.
+
+Ideally, as much of the lite client code as possible is independent of the
+particulars of pubkey/signature/blockid/vote/etc. The lite client should be
+written generically, with its own traits, and the existing types made to
+implement them. 
+
+We should follow the [lite client spec closely](https://github.com/tendermint/tendermint/blob/master/docs/spec/consensus/light-client.md), and we should work in parallel on a TLA+ implementation.
+
+Note the spec assumes a function `signers(commit)`, which returns the validators
+for a given commit. In practice, the validator set itself is not in the commit,
+so this requires fetching the validator set from a full node, computing the merkle root, 
+and comparing it against the ValidatorsHash in the header. We must also ensure 
+the hash of the header is included in the BlockID. This kind of stuff can be
+abstracted from the core lite client code (ie. `signers(commit)`), but needs to
+be supported in the actual implementation.
+
+Note the 
+[structure of commits is about to change in
+Tendermint](https://github.com/tendermint/tendermint/issues/1648),
+a big breaking change that will make blocks much smaller by eliminating
+redundant information from the included votes (ie. just including the
+signatures). The generic lite client should be able to abstract over such a
+change just fine, but we'd of course have to update how the commit types
+implement the lite client traits.
+
+We should try to surface the validity of data as much as possible in the type
+system, so we can clarify the levels of validity of our data. For instance, the
+difference between a random validator set that's just been loaded and a
+validator set that has been verified against some commits should have different
+types.
+
+We should also get started with a `lite` binary that reads from the rpc and performs a lite client sync.
+Most of the RPC client is fleshed out, but we'll have to add support for a few
+more fields and write tests.
+    
+### Reactors
+
+The primary goals of the new reactor design are [deterministic simulation](https://www.youtube.com/watch?v=4fFDFbi3toc)
+and a tight mapping between code and formal specification.
+
+Deterministic simulation should allow us to simulate networks with hundreds (thousands?!) 
+of nodes in a single process, and to deterministically replay simulations, allowing complex
+executions to be studied and debugged. The framework will require further
+investigations of Rust's concurrency model and existing frameworks.
+
+Tight mapping between code and formal specification will allow us to more easily reason
+about the correctness of the code, and, with some R&D work, automatically generate tests 
+from the formal specification. It believe this will be helped via ideas like session types,
+where as much as possible about the system's state and transitions is expressed
+in the type system.
+
+Work on both of these goals can begin independently. On the one hand, we should
+be exploring Rust frameworks for deterministic simulation, and on the other we
+should be writing the core reactor state machines, eventually to be plugged into
+the simulation framework.
+
+## Status
+
+Proposed
+
+## Consequences
+
+### Positive
+
+- Minimal changes to existing tendermint-rs
+- Parallel execution paths
+- Separation of concerns
+
+### Negative
+
+### Neutral
+
diff --git a/docs/architecture/adr-template.md b/docs/architecture/adr-template.md
new file mode 100644
index 000000000..28a5ecfbb
--- /dev/null
+++ b/docs/architecture/adr-template.md
@@ -0,0 +1,36 @@
+# ADR {ADR-NUMBER}: {TITLE}
+
+## Changelog
+* {date}: {changelog}
+
+## Context
+
+> This section contains all the context one needs to understand the current state, and why there is a problem. It should be as succinct as possible and introduce the high level idea behind the solution. 
+## Decision
+
+> This section explains all of the details of the proposed solution, including implementation details.
+It should also describe affects / corollary items that may need to be changed as a part of this.
+If the proposed change will be large, please also indicate a way to do the change to maximize ease of review.
+(e.g. the optimal split of things to do between separate PR's)
+
+## Status
+
+> A decision may be "proposed" if it hasn't been agreed upon yet, or "accepted" once it is agreed upon. If a later ADR changes or reverses a decision, it may be marked as "deprecated" or "superseded" with a reference to its replacement.
+
+{Deprecated|Proposed|Accepted}
+
+## Consequences
+
+> This section describes the consequences, after applying the decision. All consequences should be summarized here, not just the "positive" ones.
+
+### Positive
+
+### Negative
+
+### Neutral
+
+## References
+
+> Are there any relevant PR comments, issues that led up to this, or articles referrenced for why we made the given design choice? If so link them here!
+
+* {reference link}