boundflow-charter 0.1.0.dev122__tar.gz

boundflow_charter-0.1.0.dev122/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or Derivative
+          Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and do
+          not modify the License. You may add Your own attribution notices
+          within Derivative Works that You distribute, alongside or as an
+          addendum to the NOTICE text from the Work, provided that such
+          additional attribution notices cannot be construed as modifying
+          the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 Arjun Lama
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

boundflow_charter-0.1.0.dev122/PKG-INFO

@@ -0,0 +1,480 @@
+Metadata-Version: 2.4
+Name: boundflow-charter
+Version: 0.1.0.dev122
+Summary: Declarative, governed agents on BoundFlow — objective, tools, and policy as YAML.
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://github.com/boundflow/charter
+Project-URL: Source, https://github.com/boundflow/charter
+Project-URL: Issues, https://github.com/boundflow/charter/issues
+Keywords: agents,llm,deepagents,mcp,governance,control-plane
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Software Development :: Libraries :: Application Frameworks
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: boundflow>=0.6.0
+Requires-Dist: deepagents>=0.7.7
+Requires-Dist: langchain-core>=1.5
+Requires-Dist: langgraph>=1.2
+Requires-Dist: langgraph-checkpoint-postgres>=3.1
+Requires-Dist: langchain-anthropic>=0.3
+Requires-Dist: oras>=0.2
+Requires-Dist: langchain-mcp-adapters>=0.1
+Requires-Dist: mcp>=1.0
+Requires-Dist: pydantic>=2.0
+Requires-Dist: pyyaml>=6.0
+Requires-Dist: typer>=0.12
+Provides-Extra: dev
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-asyncio; extra == "dev"
+Requires-Dist: pytest-timeout; extra == "dev"
+Provides-Extra: ui
+Requires-Dist: boundflow[ui]; extra == "ui"
+Provides-Extra: otel
+Requires-Dist: boundflow[otel]; extra == "otel"
+Dynamic: license-file
+
+# Charter
+
+**The easiest way to build and manage production-ready agents.**
+
+> **Pre-alpha, and in the open early.** The design is settled enough to read and
+> argue with; the code is not settled enough to run anything you care about.
+> Expect the configuration format to change.
+
+A prototype agent is a prompt and some tools. A production agent needs more: a
+defined responsibility, explicit authority over what it may touch, a budget it
+can't exceed, a way for humans to intervene while it works, and someone watching
+its behavior over time.
+
+Charter is that, as configuration. You describe an agent in YAML — what it's
+responsible for, which tools it gets, what needs a human, what it may spend — and
+Charter deploys it as a durable, governed service you can operate.
+
+The agent runs in your environment. Its state, policy and history live in a
+control plane, so a task survives a closed laptop, a restarted worker, or an
+approval that takes until tomorrow.
+
+> [!WARNING]
+> **Pre-alpha.** Charter agents have run end-to-end against a live control plane,
+> a real model and real MCP servers — but plenty hasn't. Expect rough edges and
+> expect the configuration format to change.
+
+## Define an agent
+
+An agent is a directory with a version file in it — `refund-triage/v1.yaml`:
+
+```yaml
+apiVersion: charter/v1
+kind: AgentConfig
+
+name: refund-triage
+version: 1
+model: claude-haiku-4-5
+
+objective: |
+  Resolve the refund request on ticket {{ inputs.ticket_id }}.
+  Look up the ticket and the charge before proposing anything.
+  Never propose a refund above ${{ inputs.max_refund_usd }}.
+
+inputs:
+  ticket_id: { type: string, required: true }
+  max_refund_usd: { type: number, default: 100 }
+
+mcp:
+  - name: stripe
+    url: https://mcp.stripe.com
+    env: [STRIPE_API_KEY]
+    tools:
+      - tool: get_charge
+      - tool: create_refund
+        approval: always
+
+response_format:
+  resolution: { type: string }
+  refunded_usd: { type: number }
+```
+
+That file is the whole agent. Budgets and lifecycle rules live in two optional
+files beside it — see [a project](#a-project) — and a conservative default
+ceiling applies until you add them.
+
+Deploy it, create an instance, and give it work:
+
+```bash
+charter agent create refund-triage
+charter apply .
+charter run refund-triage --instance a3f9c012 --ticket-id 4821
+```
+
+Suppose the agent reads the ticket and the charge and concludes a $240 refund is
+warranted. It cannot issue it. `create_refund` requires approval, so Charter
+parks the task and surfaces the proposed action and the agent's reasoning to a
+human:
+
+```bash
+charter approve apr_01J8Z --reason "third dispute this month"
+```
+
+Only then is the refund executed. The agent receives the result, finishes the
+task, and reports what happened. The task didn't live in your terminal in the
+meantime.
+
+## What makes it production-ready
+
+### Explicit authority
+
+Access to an MCP server is not access to everything that server exposes. The
+model only ever sees the tools you declared:
+
+```
+mcp stripe: 34 tools available, 2 declared (32 ignored)
+```
+
+If the server adds a tool tomorrow, your agent doesn't silently gain a capability.
+And tools marked `approval: always` aren't callable inside the agent loop at all —
+the agent can propose them, but a separate step executes them after a human signs
+off. The model can propose authority it doesn't have; it can't grant it to itself.
+
+### A budget per task
+
+```yaml
+per_run:
+  max_cost_usd: 0.30
+  max_llm_calls: 40
+  max_tool_failures: 3
+```
+
+Limits hold across the whole task — reasoning rounds, retries, human feedback and
+tool calls alike. When one is exhausted, Charter tells you the operational reason
+rather than collapsing it into a generic failure:
+
+```
+stripe__create_refund failed 3 times (max_tool_failures=3)
+the integration looks broken
+```
+
+### Humans in the loop, durably
+
+An agent stops for a human for two reasons: it needs information it shouldn't
+guess, or it wants to do something beyond its authority.
+
+```bash
+charter pending refund-triage
+charter approve apr_01J8Z --reason "confirmed duplicate"
+charter answer  inp_01J8Z "use the March charge"
+```
+
+Neither is a process sitting around waiting. Charter checkpoints the task; it can
+wait overnight and resume on another worker with everything it had discovered,
+spent and been told. Rejection carries a reason back to the agent, so a "no" is
+feedback it can act on, not just a closed door.
+
+### Policy that acts on its own
+
+Humans govern individual actions. Lifecycle rules govern the agent itself:
+
+```yaml
+rules:
+  - when: { metric: num_failures, threshold: 2 }
+    then: { pause: { window: 5 } }
+
+  - when: { metric: cost, threshold: 5.00 }
+    then: { cooldown: { window: 20, seconds: 300 } }
+
+  - when: { metric: approval_rejections, threshold: 3 }
+    then: { set_version: { target: 1 } }
+```
+
+A noisy agent gets cooled down. A repeatedly failing one gets paused. A new
+version whose decisions keep getting rejected rolls back to the one that worked —
+and because versions are immutable specifications, rollback restores the whole
+agent, not just a prompt string. What changed is a file in Git, with an author and
+a diff.
+
+## Operate agents, not sessions
+
+An agent persists beyond any single task.
+
+```
+$ charter agents
+
+AGENT           VER  STATUS  ACTIVITY
+invoice-chaser  v1   paused  idle
+refund-triage   v1   active  awaiting_approval
+ticket-sweeper  v1   active  idle
+```
+
+Inspect its authority and how close it is to tripping a rule:
+
+```
+$ charter describe refund-triage
+
+limits per task
+  max_cost_usd    0.25
+  max_llm_calls   20
+
+rules
+  num_failures   1 of 2     -> pause window=5
+  cost           5.2 of 5   -> cooldown window=20 seconds=300
+```
+
+And reconstruct, afterwards, every governance decision that was made:
+
+```
+$ charter audit refund-triage
+
+2026-08-18 03:47  approval rejected by dana@example.com
+                  refund-triage: run stripe__create_refund
+                  amount_usd: 240
+                  reason: wrong charge — ch_9001 is the original
+```
+
+## Getting started
+
+1. **Define the agent** — `v1.yaml`: its objective, the tools it may call, which of
+   them need a human.
+2. **Set its policy** — `runtime.yaml` for budgets and authority, `lifecycle.yaml`
+   for pause, cooldown and rollback rules.
+3. **Package it** — `charter push` seals the version into your registry. Skip it
+   while developing: a worker reads a directory just as well.
+4. **Configure a worker** — `worker.yaml`: credentials, and which agents and
+   versions this process serves.
+5. **Apply it** — `charter apply` arms config and policy on the control plane;
+   `charter agent create` instantiates the agent.
+6. **Run it** — `charter run` starts a task; `charter status` says how it went.
+7. **Manage it** — the step that doesn't end: approve what it proposes, watch what
+   it spends, add a `v2.yaml` when it should behave differently. `charter ui` puts
+   the same thing in a browser, for whoever decides an approval.
+
+```bash
+pip install --pre boundflow-charter    # add [ui] for the console, [otel] for traces
+```
+
+`--pre` is required for now: every green build of main is published, and there is
+no stable release yet. Once one is tagged, `pip install boundflow-charter` gets it
+and `--pre` keeps meaning "whatever main is".
+
+### A control plane
+
+Charter needs one to run agents against, and there are two ways to have one:
+
+**BoundFlow Cloud** — managed, early access. Nothing to deploy, and you get two
+addresses and an API key. This is the shorter path and what the rest of this
+assumes; [request access](mailto:hello@boundflow.dev).
+
+**Self-hosted** — the BoundFlow backend is open source and runs as a container.
+Its [deployment docs](https://github.com/boundflow/boundflow/blob/main/docs/deployment.md)
+own that story; Charter only needs the addresses it gives you.
+
+Either way, inference stays yours. The control plane never sees your model key or
+its traffic.
+
+### A project
+
+```
+.
+├── worker.yaml               # which agents this worker serves, pricing, channels
+└── leads-finder/
+    ├── v1.yaml               # objective, tools, gates — versioned, immutable
+    ├── v1/skills/            # procedures for that version, shipped with it
+    ├── runtime.yaml          # budgets, limits, authority — policy, not versioned
+    └── lifecycle.yaml        # pause, cooldown and rollback rules
+```
+
+Only `v1.yaml` is required; a version file plus credentials is enough to run an
+agent. The split matters: `v1.yaml` is what the agent *does* and is versioned, so
+a rollback restores behavior. Budgets and lifecycle rules are today's guardrails
+and stay in force across one.
+
+Skills are the procedures the agent should follow once it gets there — how your
+refunds policy works, the runbook a new hire would be handed. Drop them in
+`v1/skills/<name>/SKILL.md` and they ship with that version; the layout is
+deepagents' own, so skills you already have work unchanged. They live inside `v1/`
+because rolling back to v1 should restore the instructions v1 was running with.
+
+### Traces
+
+Every model call and tool call, with its prompts, results and token counts, goes to
+a sink the worker owns:
+
+```yaml
+trace_sink:
+  kind: otel
+  endpoint: ${OTEL_EXPORTER_OTLP_ENDPOINT}
+```
+
+`otel` speaks OTLP and follows OpenTelemetry's GenAI conventions, so Jaeger, Tempo,
+Datadog and Langfuse all read it without a translation layer — `jsonl` and
+`logging` are there for when you just want to look. Traces are captured worker-side
+and carry prompts, so they go to your backend and never to the control plane.
+
+That is separate from `store.url`, which holds checkpoints and the agent's files —
+state a parked task resumes from, not telemetry.
+
+### Which agents a worker serves
+
+```yaml
+serves:
+  - agent: leads-finder       # from ./leads-finder, while you develop
+    versions: [1]
+  - agent: refund-triage      # from the registry, once it is real
+    versions: [1, 2]
+    repository: ghcr.io/acme/agents
+```
+
+A repository derives one address per version — `<repository>/<agent>:v<N>`, the
+address `charter push` writes. List every version a lifecycle rule can roll back
+to: a worker that cannot build the old version leaves the control plane
+dispatching work nobody can handle.
+
+### Credentials
+
+`worker.yaml` is committed, so nothing secret goes in it. Credentials are `${VAR}`
+references resolved from the environment when the worker starts:
+
+```yaml
+control_plane:
+  endpoint: ${BOUNDFLOW_SERVER_ADDRESS}          # the control API, for the CLI
+  worker_endpoint: ${BOUNDFLOW_WORKER_ADDRESS}   # where workers claim tasks
+  api_key: ${BOUNDFLOW_API_KEY}
+  tenant: default
+
+llm:
+  provider: anthropic
+  api_key: ${ANTHROPIC_API_KEY}
+
+store:
+  url: ${CHARTER_STORE_URL}
+```
+
+Two addresses, because BoundFlow serves the control API and worker dispatch
+separately — on two ports locally, and on two hosts in Cloud. Leave
+`worker_endpoint` out and workers fall back to `BOUNDFLOW_WORKER_ADDRESS`, then to
+localhost, which is what you want while developing and never what you want against
+a remote control plane.
+
+Inference is bring-your-own: your model key stays in the worker environment and
+model traffic never reaches the control plane. The CLI reads this same file, so
+the control plane is configured once and every command talks to the one your
+worker does.
+
+### First run
+
+```bash
+charter validate .                    # parse and cross-check every file
+charter agent create leads-finder     # bring one instance into existence
+charter apply .                       # arm config, policy and pricing
+charter worker .                      # in its own terminal — this is the process
+```
+
+`create` is separate from `apply` because an instance owns state — its own store,
+budget and lifecycle history — so bringing one into existence is a decision a
+person makes, not something CI does on their behalf. `apply` is safe to re-run as
+often as you like.
+
+`create` prints an instance id; keep it. Commands that act on an agent name an
+instance, because the instance is the thing holding the state:
+
+```bash
+charter run leads-finder --instance a3f9c012 --topic "..."
+```
+
+### Deploying
+
+Locally, `charter worker .` is the whole thing, which is what you want while
+iterating — the agent's MCP servers and your logs are right in front of you.
+
+For anything long-lived, run the worker as a container. The reason is isolation
+rather than packaging: an agent can declare MCP servers as commands, and the
+worker executes them. [deploy/](deploy/) has a Dockerfile that mounts the project
+rather than baking it in, so changing an objective is a restart, not a rebuild.
+
+## CLI
+
+Working with configuration:
+
+```bash
+charter validate .               # parse and cross-check configuration
+charter diff .                   # compare declared and deployed state
+charter apply .                  # update config, policy and pricing
+charter push <agent> <ref>       # publish a version to an OCI registry
+charter worker .                 # run agents in this environment
+```
+
+Operating an agent:
+
+```bash
+charter agent create <agent>     # bring an instance into existence
+charter run <agent> --flags      # start a task
+charter agents                   # every agent and what it's doing
+charter describe <agent>         # authority, limits, rules, any hold
+charter tasks <agent>            # task history
+charter status <task-id>         # result, cost and why it stopped
+charter audit <agent>            # every governance decision recorded
+
+charter pending <agent>          # the open gate, if it's parked on one
+charter approve / reject / answer <id>
+charter ui                       # the same, in a browser
+
+charter pause <agent>            # stop it taking work
+charter resume <agent> --suspension <id>
+charter agent delete <agent>     # destroy an instance and its history
+```
+
+Read commands take `--json` for the complete record instead of the curated view.
+
+## Architecture
+
+Charter is an opinionated agent layer built on
+[BoundFlow](https://github.com/boundflow/boundflow).
+
+`charter apply` compiles your configuration into workflows and policy on the
+BoundFlow control plane. Charter workers run the actual agent loop in your
+environment, talking to your MCP servers with credentials that never leave it.
+
+```
+                    BoundFlow
+                  Control Plane
+             state • policy • lifecycle
+                       │
+                      RPC
+                       │
+                       ▼
+              Your environment
+        ┌─────────────────────────┐
+        │ Charter worker          │
+        │                         │
+        │ model ↔ agent loop      │
+        │             │           │
+        │          MCP tools      │
+        └─────────────────────────┘
+```
+
+Charter introduces no database or service of its own. If the Charter CLI vanished,
+deployed agents would keep running through their workers and the control plane.
+
+See [DESIGN.md](DESIGN.md) for the full configuration reference and the design
+decisions behind it, and [examples/](examples/) for complete configurations.
+
+## Development
+
+```bash
+python -m venv .venv
+.venv/bin/pip install -e ../boundflow/sdk/python   # your BoundFlow checkout
+.venv/bin/pip install -e '.[dev]'
+.venv/bin/pytest
+```
+
+End-to-end tests require a BoundFlow control plane and are excluded by default:
+
+```bash
+docker compose -f ../boundflow/docker-compose.dist.yml up -d
+export BOUNDFLOW_API_KEY=<...>
+pytest tests/e2e
+```
+
+They use a real control plane, a real MCP subprocess and real governance gates.
+Only the model is faked, so the suite stays deterministic and free.